JPWO2007069350A1 - Image encoding and decoding method and apparatus - Google Patents

Abstract

The encoding technique includes a step of performing matching calculation between both end image frames of an image group including three or more image frames, and a corresponding end point image based on corresponding point information between both end image frames obtained as a result of the matching calculation. The step of virtually generating an intermediate image frame sandwiched between frames by interpolation, and whether any of the virtually generated intermediate image frames has a difference greater than or equal to an allowable value from the actual intermediate image frame Determining whether or not there is an intermediate image frame having a difference greater than or equal to an allowable value, and identifying a region having a large difference on the image of the intermediate image frame; Generating encoded data including difference information regarding the identified region, at least one of the two-end image frames, and corresponding point information. And a step of.

Description

  The present invention relates to encoding and decoding techniques for images, particularly moving images.

MPEG (Motion Picture Experts Group) is one standard technology for video compression. In MPEG, block matching is used. In this matching, a block search is performed so that the difference between blocks is minimized. For this reason, the difference is surely small, but areas that originally correspond to each other between frames are not necessarily associated with each other.
Patent No. 2927350

  In MPEG, so-called block noise becomes a problem when trying to increase the compression rate. In order to suppress the generation of this noise and further increase the compression ratio focused on inter-frame coherency, it is necessary to modify the current block matching-based technology. The technique to be obtained should be encoded so that the originally corresponding areas or pixels correspond correctly, and it is desirable to avoid simple block matching.

  The object of the present invention is as follows. First, a moving picture compression technique that does not cause block noise that is a problem in MPEG, that is, a moving picture coding technique is provided. Also provided is a moving picture decoding technique corresponding to the moving picture encoding technique. As another object, a new moving picture encoding and decoding technique using an image matching technique different from MPEG is provided. Still another object is to provide different image encoding and decoding techniques as a whole while using an image matching technique similar to MPEG.

  In one aspect of the image encoding method of the present invention, an intermediate image frame is generated by interpolation calculation based on corresponding point information between the first and second image frames. When the difference between the intermediate image frame and the actually existing intermediate image frame is large, an area where the difference is actually large is specified in the image. Next, encoded data is generated in a form including difference information regarding the identified region, at least data of the first or second image frame, and corresponding point information. Decoding techniques follow the reverse process.

  It is also effective as the present invention to exchange the above steps, to exchange or add part or all of the expression between the method and the apparatus, or to change the expression to a computer program, a recording medium, or the like.

  According to the present invention, an effect corresponding to the above object can be obtained.

1 (a) and 1 (b) are images obtained by applying an averaging filter to the faces of two persons, and FIGS. 1 (c) and 1 (d) are prerequisite technologies relating to the faces of the two persons. 1 (e) and FIG. 1 (f) are images of p (5,1) obtained by the base technology regarding the faces of two persons, FIG. 1 (g). And FIG. 1 (h) is an image of p (5, 2) required by the base technology regarding the faces of the two persons, and FIGS. 1 (i) and 1 (j) are base technologies regarding the faces of the two persons. It is the photograph of the halftone image which each displayed the image of calculated | required p (5, 3) on a display. FIG. 2 (R) is a diagram showing the original quadrilateral, and FIG. 2 (A), FIG. 2 (B), FIG. 2 (C), FIG. 2 (D), and FIG. FIG. It is a figure which shows the relationship between a starting point image and an end point image, and the relationship between the m-th level and the m-1st level using an inheritance quadrilateral. It is a figure which shows the relationship between parameter (eta) and energy Cf. FIG. 5A and FIG. 5B are diagrams illustrating a state in which whether or not a mapping related to a certain point satisfies the bijection condition is calculated from outer product calculation. It is a flowchart which shows the whole procedure of a premise technique. It is a flowchart which shows the detail of S1 of FIG. It is a flowchart which shows the detail of S10 of FIG. It is a figure which shows the correspondence of a part of image of a m-th level, and a part of image of a m-1st level. It is a figure which shows the starting point hierarchy image produced | generated by the base technology. It is a figure which shows the procedure of the preparation of matching evaluation before progressing to S2 of FIG. It is a flowchart which shows the detail of S2 of FIG. It is a figure which shows a mode that a submapping is determined in the 0th level. It is a figure which shows a mode that a submapping is determined in a 1st level. It is a flowchart which shows the detail of S21 of FIG. It is a figure which shows the behavior of energy C (m, s) f corresponding to f (m, s) (λ = iΔλ) obtained while changing λ for a certain f (m, s). It is a figure which shows the behavior of energy C (n) f corresponding to f (n) (η = iΔη) (i = 0, 1,...) obtained while changing η. It is a flowchart which calculates | requires the submapping in the m-th level in the premise technique after improvement. It is a figure which shows the structure of the flow of the image coding technique which concerns on 1st Embodiment, and an image coding apparatus. 20A to 20C are diagrams illustrating examples of target image frames. It is a figure which shows the data format of the image coding technique which concerns on 1st Embodiment. It is a figure which shows the structure of the flow of the image decoding technique which concerns on 1st Embodiment, and an image decoding apparatus. It is a figure which shows the flow of the image decoding technique which concerns on 2nd Embodiment, and the structure of an image decoding apparatus. It is a figure which shows the flow of the image decoding technique which concerns on 2nd Embodiment, and the structure of an image decoding apparatus. It is a figure which shows the structure of DE + NR of FIG. 23 which concerns on embodiment.

  In the following embodiment, an image matching technique is used. For this technique, the technique previously proposed by the present applicant in Japanese Patent No. 2927350 (hereinafter referred to as “premise technique”) can be used. However, other matching techniques may be used. Hereinafter, in any aspect, the modifications and considerations described in any part may be applied to the same technique in another part.

  First, the multi-resolution singularity filter technique used in the embodiment and the image matching process using the technique will be described in detail as a “premise technique”.

[Embodiment of prerequisite technology]
First, the elemental technology of the prerequisite technology is described in detail in [1], and the processing procedure is specifically described in [2]. In [3], the point of improvement based on the prerequisite technology is described.
[1] Details of element technology [1.1] Introduction A new multi-resolution filter called a singularity filter is introduced to accurately calculate matching between images. No prior knowledge of objects is required. The calculation of matching between images is calculated at each resolution while proceeding through the resolution hierarchy. At that time, the resolution hierarchy is sequentially traced from the coarse level to the fine level. The parameters required for the calculation are set completely automatically by dynamic calculations similar to the human visual system. There is no need to manually identify corresponding points between images.

  The base technology can be applied to, for example, fully automatic morphing, object recognition, stereoscopic photogrammetry, volume rendering, and generation of a smooth moving image from a small number of frames. When used for morphing, a given image can be automatically transformed. When used for volume rendering, an intermediate image between cross sections can be accurately reconstructed. The same applies to the case where the distance between the cross sections is long and the shape of the cross section changes greatly.

[1.2] Hierarchy of Singularity Filter The multi-resolution singularity filter according to the base technology can preserve the luminance and position of each singularity included in the image while reducing the resolution of the image. Here, the width of the image is N, and the height is M. Hereinafter, for simplicity, it is assumed that N = M = 2n (n is a natural number). The section [0, N] ＮR is described as I. The pixel of the image at (i, j) is described as p (i, j) (i, jεI).

ただしここで、

とする。以降これら４つの画像を副画像（サブイメージ）と呼ぶ。ｍｉｎｘ≦ｔ≦ｘ＋１、ｍａｘｘ≦ｔ≦ｘ＋１をそれぞれα及びβと記述すると、副画像はそれぞれ以下のように記述できる。Here, a multi-resolution hierarchy is introduced. The hierarchized image group is generated by a multi-resolution filter. The multi-resolution filter performs a two-dimensional search on the original image to detect singular points, and generates another image having a lower resolution than the original image even if the detected singular points are extracted. . Here, the size of each image at the m-th level is 2m × 2m (0 ≦ m ≦ n). The singularity filter recursively constructs the following four types of new hierarchical images in a direction descending from n.

Where

And Hereinafter, these four images are called sub-images. If minx ≦ t ≦ x + 1 and maxx ≦ t ≦ x + 1 are described as α and β, respectively, the sub-images can be described as follows.

P (m, 0) = α (x) α (y) p (m + 1,0)
P (m, 1) = α (x) β (y) p (m + 1,1)
P (m, 2) = β (x) α (y) p (m + 1, 2)
P (m, 3) = β (x) β (y) p (m + 1,3)
That is, these are considered to be like tensor products of α and β. Each sub-image corresponds to a singular point. As is clear from these equations, the singularity filter detects a singularity for each block composed of 2 × 2 pixels in the original image. At this time, a point having the maximum pixel value or the minimum pixel value is searched for in two directions of each block, that is, vertical and horizontal. As the pixel value, luminance is adopted in the base technology, but various numerical values relating to the image can be adopted. The pixel with the maximum pixel value in both directions is the maximum point, the pixel with the minimum pixel value in both directions is the minimum point, the maximum pixel value in one of the two directions, and the minimum pixel value in the other direction Are detected as saddle points.

  The singularity filter reduces the resolution of the image by representing the image (4 pixels here) of the block with the image (1 pixel here) of the singularity detected inside each block. From the theoretical point of view of singularities, α (x) α (y) preserves the minimum point, β (x) β (y) preserves the maximum point, α (x) β (y) and β (x) α (y) preserves saddle points.

  First, the singular point filtering process is separately performed on the start point (source) image and the end point (destination) image to be matched, and a series of images, that is, a start point hierarchical image and an end point hierarchical image are generated. Four types of start point hierarchical images and end point hierarchical images are generated corresponding to the types of singular points.

  Thereafter, matching between the start point layer image and the end point layer image is performed within a series of resolution levels. First, the minimum point is matched using p (m, 0). Next, based on the result, matching of saddle points is performed using p (m, 1), and matching of other saddle points is performed using p (m, 2). Finally, the maximum points are matched using p (m, 3).

  FIGS. 1C and 1D show the sub-image p (5, 0) of FIGS. 1A and 1B, respectively. Similarly, FIG. 1 (e) and FIG. 1 (f) are p (5, 1), FIG. 1 (g) and FIG. 1 (h) are p (5, 2), FIG. 1 (i) and FIG. j) represents p (5, 3), respectively. As can be seen from these figures, the sub-image facilitates the matching of the feature portions of the image. First, the eye becomes clear by p (5,0). This is because the eyes are the minimum brightness points in the face. According to p (5,1), the mouth becomes clear. This is because the mouth is low in luminance in the horizontal direction. According to p (5, 2), the vertical lines on both sides of the neck become clear. Finally, p (5,3) makes the brightest point of the ear and cheek clear. This is because these are the maximum points of luminance.

  The feature of the image can be extracted by using the singular point filter. For example, by comparing the features of the image captured by the camera with the features of several objects that have been recorded in advance, the subject reflected in the camera is identified. can do.

[1.3] Calculation of mapping between images The pixel at the position (i, j) of the start point image is written as p (n) (i, j), and the pixel at the position (k, l) of the end point image is also q ( n) Described by (k, l). Let i, j, k, lεI. Defines the energy of mapping between images (described later). This energy is determined by the difference between the luminance of the pixel of the start point image and the luminance of the corresponding pixel of the end point image, and the smoothness of the mapping. First, a mapping f (m, 0): p (m, 0) → q (m, 0) between p (m, 0) and q (m, 0) having the minimum energy is calculated. Based on f (m, 0), a mapping f (m, 1) between p (m, 1) and q (m, 1) having the minimum energy is calculated. This procedure continues until the computation of the mapping f (m, 3) between p (m, 3) and q (m, 3) is completed. Each map f (m, i) (i = 0, 1, 2,...) Is called a sub-map. For convenience of calculation of f (m, i), the order of i can be rearranged as follows: The reason why the rearrangement is necessary will be described later.

ここでσ（ｉ）∈｛０，１，２，３｝である。

Here, σ (i) ε {0, 1, 2, 3}.

[1.3.1] Bijection When the matching between the start point image and the end point image is expressed by mapping, the mapping should satisfy the bijection condition between the two images. This is because there is no conceptual superiority or inferiority between the two images, and the pixels of each other should be connected both surjectively and injectively. However, unlike the normal case, the map to be constructed here is a bijective digital version. In the base technology, a pixel is specified by a grid point.

  The mapping from the start point sub-image (sub-image provided for the start-point image) to the end-point sub-image (sub-image provided for the end-point image) is f (m, s): I / 2n−m × I / 2n−m → I / 2n−m × I / 2n−m (s = 0, 1,...) Here, f (m, s) (i, j) = (k, l) indicates that p (m, s) (i, j) of the start point image is q (m, s) (k, l) of the end point image. ). For simplicity, the pixel q (k, l) is described as qf (i, j) when f (i, j) = (k, l) holds.

を考える（ｉ＝０，…，２ｍ−１、ｊ＝０，…，２ｍ−１）。ここでＲの各辺（エッジ）の方向を以下のように定める。

この正方形は写像ｆによって終点画像平面における四辺形に写像されなければならない。ｆ（ｍ，ｓ）（Ｒ）によって示される四辺形、

は、以下の全単射条件を満たす必要がある。The definition of bijection is important when data is discrete, such as pixels (grid points) handled in the base technology. Here, it is defined as follows (i, i ', j, j', k, l are all integers). First, each square region denoted by R in the plane of the starting point image,

(I = 0,..., 2m-1, j = 0,..., 2m-1). Here, the direction of each side (edge) of R is determined as follows.

This square must be mapped to a quadrilateral in the endpoint image plane by mapping f. a quadrilateral indicated by f (m, s) (R),

Must meet the following bijective conditions:

1. The edges of the quadrilateral f (m, s) (R) do not intersect each other.
2. The direction of the edges of f (m, s) (R) is equal to those of R (clockwise in the case of FIG. 2).
3. As a relaxation condition, retraction is allowed.

  This is because unless there is any relaxation condition, the only map that completely satisfies the bijection condition is the unit map. Here, the length of one edge of f (m, s) (R) is 0, that is, f (m, s) (R) may be a triangle. However, it should not be a figure with an area of 0, that is, one point or one line segment. When FIG. 2 (R) is the original quadrilateral, FIG. 2 (A) and FIG. 2 (D) satisfy the bijection condition, but FIG. 2 (B), FIG. 2 (C), FIG. ) Is not satisfied.

  In an actual implementation, the following conditions may be further imposed to easily guarantee that the mapping is surjective. That is, each pixel on the boundary of the start point image is mapped to a pixel occupying the same position in the end point image. That is, f (i, j) = (i, j) (where, i = 0, i = 2m-1, j = 0, j = 2m-1 on four lines). Hereinafter, this condition is also referred to as “additional condition”.

ここで、Ｖ（ｐ（ｍ，ｓ）（ｉ，ｊ））及びＶ（ｑ（ｍ，ｓ）ｆ（ｉ，ｊ））はそれぞれ画素ｐ（ｍ，ｓ）（ｉ，ｊ）及びｑ（ｍ，ｓ）ｆ（ｉ，ｊ）の輝度である。ｆのトータルのエネルギーＣ（ｍ，ｓ）は、マッチングを評価するひとつの評価式であり、つぎに示すＣ（ｍ，ｓ）（ｉ，ｊ）の合計で定義できる。

［１．３．２．２］滑らかな写像のための画素の位置に関するコスト
滑らかな写像を得るために、写像に関する別のエネルギーＤｆを導入する。このエネルギーは画素の輝度とは関係なく、ｐ（ｍ，ｓ）（ｉ，ｊ）およびｑ（ｍ，ｓ）ｆ（ｉ，ｊ）の位置によって決まる（ｉ＝０，…，２ｍ−１，ｊ＝０，…，２ｍ−１）。点（ｉ，ｊ）における写像ｆ（ｍ，ｓ）のエネルギーＤ（ｍ，ｓ）（ｉ，ｊ）は次式で定義される。

ただし、係数パラメータηは０以上の実数であり、また、

とする。ここで、

であり、ｉ’＜０およびｊ’＜０に対してｆ（ｉ’，ｊ’）は０と決める。Ｅ０は（ｉ，ｊ）及びｆ（ｉ，ｊ）の距離で決まる。Ｅ０は画素があまりにも離れた画素へ写影されることを防ぐ。ただしＥ０は、後に別のエネルギー関数で置き換える。Ｅ１は写像の滑らかさを保証する。Ｅ１は、ｐ（ｉ，ｊ）の変位とその隣接点の変位の間の隔たりを表す。以上の考察をもとに、マッチングを評価する別の評価式であるエネルギーＤｆは次式で定まる。

［１．３．２．３］写像の総エネルギー
写像の総エネルギー、すなわち複数の評価式の統合に係る総合評価式はλＣ（ｍ，ｓ）ｆ＋Ｄ（ｍ，ｓ）ｆで定義される。ここで係数パラメータλは０以上の実数である。目的は総合評価式が極値をとる状態を検出すること、すなわち次式で示す最小エネルギーを与える写像を見いだすことである。

λ＝０及びη＝０の場合、写像は単位写像になることに注意すべきである（すなわち、全てのｉ＝０，…，２ｍ−１及びｊ＝０，…，２ｍ−１に対してｆ（ｍ，ｓ）（ｉ，ｊ）＝（ｉ，ｊ）となる）。後述のごとく、本前提技術では最初にλ＝０及びη＝０の場合を評価するため、写像を単位写像から徐々に変形していくことができる。仮に総合評価式のλの位置を変えてＣ（ｍ，ｓ）ｆ＋λＤ（ｍ，ｓ）ｆと定義したとすれば、λ＝０及びη＝０の場合に総合評価式がＣ（ｍ，ｓ）ｆだけになり、本来何等関連のない画素どうしが単に輝度が近いというだけで対応づけられ、写像が無意味なものになる。そうした無意味な写像をもとに写像を変形していってもまったく意味をなさない。このため、単位写像が評価の開始時点で最良の写像として選択されるよう係数パラメータの与えかたが配慮されている。[1.3.2] Mapping energy [1.3.2.1] Pixel luminance cost Defines the energy of mapping f. The purpose is to find a map that minimizes energy. The energy is mainly determined by the difference between the luminance of the pixel in the start point image and the luminance of the corresponding pixel in the end point image. That is, the energy C (m, s) (i, j) at the point (i, j) of the map f (m, s) is determined by the following equation.

Where V (p (m, s) (i, j)) and V (q (m, s) f (i, j)) are pixels p (m, s) (i, j) and q ( m, s) is the luminance of f (i, j). The total energy C (m, s) of f is one evaluation formula for evaluating matching, and can be defined by the sum of C (m, s) (i, j) shown below.

[1.3.2.2] Cost of pixel location for smooth mapping In order to obtain a smooth mapping, another energy Df for mapping is introduced. This energy is independent of the luminance of the pixel and is determined by the positions of p (m, s) (i, j) and q (m, s) f (i, j) (i = 0,..., 2m−1, j = 0,..., 2m-1). The energy D (m, s) (i, j) of the map f (m, s) at the point (i, j) is defined by the following equation.

However, the coefficient parameter η is a real number of 0 or more, and

And here,

And f (i ′, j ′) is determined to be 0 for i ′ <0 and j ′ <0. E0 is determined by the distance between (i, j) and f (i, j). E0 prevents the pixel from being mapped to a pixel that is too far away. However, E0 is replaced later with another energy function. E1 guarantees the smoothness of the mapping. E1 represents the distance between the displacement of p (i, j) and the displacement of its neighboring points. Based on the above consideration, energy Df, which is another evaluation formula for evaluating matching, is determined by the following formula.

[1.3.2.3] Total energy of mapping The total energy of mapping, that is, a comprehensive evaluation formula relating to the integration of a plurality of evaluation formulas is defined by λC (m, s) f + D (m, s) f. Here, the coefficient parameter λ is a real number of 0 or more. The purpose is to detect a state in which the comprehensive evaluation formula takes an extreme value, that is, to find a map that gives the minimum energy shown by the following formula.

Note that for λ = 0 and η = 0, the mapping is a unit mapping (ie for all i = 0,..., 2m−1 and j = 0,..., 2m−1). f (m, s) (i, j) = (i, j)). As will be described later, since the base technology first evaluates the case of λ = 0 and η = 0, the mapping can be gradually changed from the unit mapping. If the position of λ in the comprehensive evaluation formula is changed and defined as C (m, s) f + λD (m, s) f, then the comprehensive evaluation formula is C (m, s) when λ = 0 and η = 0. ) F, and the pixels that are not related to each other are associated with each other simply because the luminance is close, and the mapping becomes meaningless. Even if the map is transformed based on such a meaningless map, it does not make any sense. For this reason, consideration is given to how the coefficient parameter is given so that the unit map is selected as the best map at the start of evaluation.

  The optical flow also considers the difference in pixel brightness and the smoothness as in this base technology. However, the optical flow cannot be used for image conversion. This is because only local movement of the object is considered. A global correspondence can be detected by using a singularity filter according to the base technology.

[1.3.3] Determination of mapping by introduction of multi-resolution The minimum energy is given, and the mapping fmin that satisfies the bijection condition is obtained using a multi-resolution hierarchy. The mapping between the start sub-image and the end sub-image is calculated at each resolution level. Starting from the top of the resolution hierarchy (the coarsest level), the mapping of each resolution level is determined taking into account the mappings of the other levels. The number of mapping candidates at each level is limited by using higher or coarser level mappings. More specifically, when determining a mapping at a certain level, a mapping obtained at one coarser level is imposed as a kind of constraint condition.

が成り立つとき、ｐ（ｍ−１，ｓ）（ｉ’，ｊ’）、ｑ（ｍ−１，ｓ）（ｉ’，ｊ’）をそれぞれｐ（ｍ，ｓ）（ｉ，ｊ）、ｑ（ｍ，ｓ）（ｉ，ｊ）のｐａｒｅｎｔと呼ぶことにする。［ｘ］はｘを越えない最大整数である。またｐ（ｍ，ｓ）（ｉ，ｊ）、ｑ（ｍ，ｓ）（ｉ，ｊ）をそれぞれｐ（ｍ−１，ｓ）（ｉ’，ｊ’）、ｑ（ｍ−１，ｓ）（ｉ’，ｊ’）のｃｈｉｌｄと呼ぶ。関数ｐａｒｅｎｔ（ｉ，ｊ）は次式で定義される。

ｐ（ｍ，ｓ）（ｉ，ｊ）とｑ（ｍ，ｓ）（ｋ，ｌ）の間の写像ｆ（ｍ，ｓ）は、エネルギー計算を行って最小になったものを見つけることで決定される。ｆ（ｍ，ｓ）（ｉ，ｊ）＝（ｋ，ｌ）の値はｆ（ｍ−１，ｓ）（ｍ＝１，２，…，ｎ）を用いることによって、以下のように決定される。まず、ｑ（ｍ，ｓ）（ｋ，ｌ）は次の四辺形の内部になければならないという条件を課し、全単射条件を満たす写像のうち現実性の高いものを絞り込む。

ただしここで、

である。こうして定めた四辺形を、以下ｐ（ｍ，ｓ）（ｉ，ｊ）の相続（inherited）四辺形と呼ぶことにする。相続四辺形の内部において、エネルギーを最小にする画素を求める。First,

P (m−1, s) (i ′, j ′) and q (m−1, s) (i ′, j ′) are converted to p (m, s) (i, j) and q, respectively. This is called parent of (m, s) (i, j). [X] is a maximum integer not exceeding x. Further, p (m, s) (i, j) and q (m, s) (i, j) are respectively changed to p (m-1, s) (i ', j'), q (m-1, s). It is called a child of (i ′, j ′). The function parent (i, j) is defined by the following equation.

The mapping f (m, s) between p (m, s) (i, j) and q (m, s) (k, l) is determined by performing an energy calculation and finding the smallest one. Is done. The value of f (m, s) (i, j) = (k, l) is determined as follows by using f (m−1, s) (m = 1, 2,..., n). The First, q (m, s) (k, l) imposes a condition that it must be inside the following quadrilateral, and narrows down highly realistic maps that satisfy the bijection condition.

Where

It is. The quadrilateral thus determined is hereinafter referred to as an inherited quadrilateral of p (m, s) (i, j). Find the pixel that minimizes the energy inside the inherited quadrilateral.

  FIG. 3 shows the above procedure. In the figure, the A, B, C, and D pixels of the start point image are mapped to the end point images A ', B', C ', and D', respectively, at the (m-1) th level. Pixel p (m, s) (i, j) is mapped to pixel q (m, s) f (m) (i, j) existing inside inherited quadrangle A′B′C′D ′. There must be. With the above consideration, the mapping from the (m-1) th level mapping to the mth level mapping is performed.

また、副写像ｆ（ｍ，ｓ）を計算するためには次式を用いる。

こうしてすべての副写像のエネルギーを低い値に保つ写像が得られる。式２０により、異なる特異点に対応する副写像が、副写像どうしの類似度が高くなるように同一レベル内で関連づけられる。式１９は、ｆ（ｍ，ｓ）（ｉ，ｊ）と、第ｍ−１レベルの画素の一部と考えた場合の（ｉ，ｊ）が射影されるべき点の位置との距離を示している。The previously defined energy E0 is replaced by the following equation in order to calculate the submapping f (m, 0) at the mth level.

Further, the following formula is used to calculate the submapping f (m, s).

This gives a map that keeps the energy of all submaps low. According to Equation 20, the sub-maps corresponding to different singular points are related within the same level so that the similarity between the sub-maps becomes high. Equation 19 shows the distance between f (m, s) (i, j) and the position of the point where (i, j) should be projected when considered as a part of the pixel of the (m-1) th level. ing.

  If there is no pixel satisfying the bijection condition inside the inherited quadrilateral A'B'C'D ', the following measures are taken. First, a pixel whose distance from the boundary line of A′B′C′D ′ is L (initially L = 1) is examined. If the one with the minimum energy satisfies the bijection condition, it is selected as the value of f (m, s) (i, j). L is increased until such a point is found or L reaches its upper limit L (m) max. L (m) max is fixed for each level m. If such a point is not found at all, a mapping in which the area of the quadrilateral to be converted becomes zero by temporarily ignoring the third condition of bijection is recognized, and f (m, s) (i , J). If a point that satisfies the condition is still not found, the bijection first and second conditions are then removed.

  An approximation method using multiple resolutions is essential to determine the global correspondence between images while avoiding the mapping being affected by image details. Unless an approximation method based on multi-resolution is used, it is impossible to find a correspondence between distant pixels. In that case, the size of the image must be limited to a very small size, and only an image with small change can be handled. Furthermore, since the normal mapping requires smoothness, it is difficult to find the correspondence between such pixels. This is because the energy of mapping from pixel to pixel at a distance is high. According to the approximation method using multi-resolution, an appropriate correspondence between such pixels can be found. This is because these distances are small at the upper level (coarse level) of the resolution hierarchy.

[1.4] Automatic determination of optimal parameter values One of the main drawbacks of existing matching techniques is the difficulty in adjusting parameters. In most cases, the parameter adjustment is performed manually, and it is extremely difficult to select an optimum value. According to the method according to the base technology, the optimum parameter value can be determined completely automatically.

  The system according to the base technology includes two parameters, λ and η. In short, λ is the weight of the difference in luminance of the pixels, and η is the stiffness of the mapping. The initial values of these parameters are 0. First, η is fixed at η = 0 and λ is gradually increased from 0. When the value of the overall evaluation formula (Equation 14) is minimized while increasing the value of λ, the value of C (m, s) f for each submapping generally decreases. This basically means that the two images must be better matched. However, when λ exceeds the optimum value, the following phenomenon occurs.

1. Pixels that should not be associated with each other are erroneously associated with each other simply because the luminance is close.
2. As a result, the correspondence between the pixels becomes strange and the mapping starts to be broken.

3. As a result, D (m, s) f in Equation 14 tends to increase rapidly.
4). As a result, since the value of Equation 14 tends to increase rapidly, f (m, s) changes to suppress the rapid increase in D (m, s) f, and as a result, C (m, s) f. Will increase.

  Therefore, a threshold value at which C (m, s) f changes from decreasing to increasing is detected while maintaining a state where Equation 14 takes the minimum value while increasing λ, and λ is set as an optimum value at η = 0. Next, η is increased little by little to check the behavior of C (m, s) f, and η is automatically determined by the method described later. Λ is also determined corresponding to η.

  This method is similar to the operation of the focus mechanism of the human visual system. In the human visual system, the left and right eyes are matched while moving one eye. When an object can be clearly recognized, its eyes are fixed.

[1.4.1] Dynamic determination of λ λ is increased from 0 by a predetermined step size, and the submapping is evaluated each time the value of λ changes. As in Equation 14, the total energy is defined by λC (m, s) f + D (m, s) f. D (m, s) f in Equation 9 represents smoothness, theoretically becomes the minimum in the case of unit mapping, and E0 and E1 increase as the mapping is distorted. Since E1 is an integer, the minimum step size of D (m, s) f is 1. For this reason, if the current change (decrease amount) of λC (m, s) (i, j) is not greater than 1, the total energy cannot be reduced by changing the mapping. This is because D (m, s) f increases by 1 or more as the mapping changes, so that the total energy does not decrease unless λC (m, s) (i, j) decreases by 1 or more.

で示されるＡ個の画素が、

のエネルギーを持つより安定的な状態に変化する。ここでは仮に、これらの画素のエネルギーがすべてゼロになると近似している。この式はＣ（ｍ，ｓ）ｆの値が、

だけ変化することを示し、その結果、

が成立する。ｈ（ｌ）＞０であるから、通常Ｃ（ｍ，ｓ）ｆは減少する。しかし、λが最適値を越えようとするとき、上述の現象、つまりＣ（ｍ，ｓ）ｆの増加が発生する。この現象を検出することにより、λの最適値を決定する。Under this condition, it is shown that C (m, s) (i, j) decreases in the normal state as λ increases. A histogram of C (m, s) (i, j) is described as h (l). h (l) is the number of pixels whose energy C (m, s) (i, j) is l2. Since λl2 ≧ 1 holds, for example, consider the case of l2 = 1 / λ. When λ changes by a small amount from λ1 to λ2,

A pixels indicated by

Changes to a more stable state with the energy of. Here, it is approximated that the energy of these pixels is all zero. In this equation, the value of C (m, s) f is

Show that only changes, and as a result,

Is established. Since h (l)> 0, C (m, s) f usually decreases. However, when λ is about to exceed the optimum value, the above phenomenon, that is, an increase in C (m, s) f occurs. By detecting this phenomenon, the optimum value of λ is determined.

と仮定すれば、

が成り立つ。このときｋ≠−３であれば、

となる。これがＣ（ｍ，ｓ）ｆの一般式である（Ｃは定数）。When H (h> 0) and k are constants,

Assuming

Holds. At this time, if k ≠ -3,

It becomes. This is a general formula of C (m, s) f (C is a constant).

が成立しているため、全単射条件を破る画素の数は次式の率で増加する。

従って、

は定数である。仮にｈ（ｌ）＝Ｈｌｋを仮定するとき、例えば、

は定数になる。しかしλが最適値を越えると、上の値は急速に増加する。この現象を検出し、Ｂ０λ３／２＋ｋ／２／２ｍの値が異常値Ｂ０ｔｈｒｅｓを越えるかどうかを検査し、λの最適値を決定することができる。同様に、Ｂ１λ３／２＋ｋ／２／２ｍの値が異常値Ｂ１ｔｈｒｅｓを越えるかどうかを検査することにより、全単射の第３の条件を破る画素の増加率Ｂ１を確認する。ファクター２ｍを導入する理由は後述する。このシステムはこれら２つの閾値に敏感ではない。これらの閾値は、エネルギーＣ（ｍ，ｓ）ｆの観察では検出し損なった写像の過度の歪みを検出するために用いることができる。When detecting the optimum value of λ, the number of pixels that violate the bijection condition may be inspected for further safety. Here, when determining the mapping of each pixel, it is assumed that the probability of breaking the bijection condition is p0. in this case,

Therefore, the number of pixels that violate the bijection condition increases at the rate of the following equation.

Therefore,

Is a constant. If h (l) = Hlk is assumed, for example,

Becomes a constant. However, when λ exceeds the optimum value, the above value increases rapidly. By detecting this phenomenon, it is possible to determine whether the value of B0λ3 / 2 + k / 2 / 2m exceeds the abnormal value B0thres and determine the optimum value of λ. Similarly, by checking whether the value of B1λ3 / 2 + k / 2 / 2m exceeds the abnormal value B1thres, the pixel increase rate B1 that violates the third condition of bijection is confirmed. The reason for introducing the factor 2m will be described later. This system is not sensitive to these two thresholds. These thresholds can be used to detect excessive distortion of the mapping that was missed by observation of energy C (m, s) f.

  In the experiment, when calculating the submapping f (m, s), if λ exceeds 0.1, the calculation of f (m, s) is stopped and the calculation shifts to the calculation of f (m, s + 1). This is because a difference of only “3” in the luminance level of 255 pixels affects the submapping calculation when λ> 0.1, and it is difficult to obtain a correct result when λ> 0.1. .

[1.4.2] Histogram h (l)
The examination of C (m, s) f does not depend on the histogram h (l). It can be affected by h (l) during inspection of bijection and its third condition. When (λ, C (m, s) f) is actually plotted, k is usually near 1. In the experiment, k = 1 was used, and B0λ2 and B1λ2 were examined. If the true value of k is less than 1, B0λ2 and B1λ2 do not become constants, but gradually increase according to the factor λ (1-k) / 2. If h (l) is a constant, for example, the factor is λ1 / 2. However, these differences can be absorbed by setting the threshold value B0thres correctly.

一方、終点画像は、次式のごとく中心（ｘ１，ｙ１）、半径がｒのオブジェクトであるとする。

ここでｃ（ｘ）はｃ（ｘ）＝ｘｋの形であるとする。中心（ｘ０，ｙ０）及び（ｘ１，ｙ１）が十分遠い場合、ヒストグラムｈ（ｌ）は次式の形となる。

ｋ＝１のとき、画像は背景に埋め込まれた鮮明な境界線を持つオブジェクトを示す。このオブジェクトは中心が暗く、周囲にいくに従って明るくなる。ｋ＝−１のとき、画像は曖昧な境界線を持つオブジェクトを表す。このオブジェクトは中心が最も明るく、周囲にいくに従って暗くなる。一般のオブジェクトはこれらふたつのタイプのオブジェクトの中間にあると考えてもさして一般性を失わない。したがって、ｋは−１≦ｋ≦１として大抵の場合をカバーでき、式２７が一般に減少関数であることが保障される。Here, it is assumed that the starting point image is a circular object having a center at (x0, y0) and a radius r as in the following equation.

On the other hand, it is assumed that the end point image is an object having a center (x1, y1) and a radius r as in the following equation.

Here, c (x) is assumed to be in the form of c (x) = xk. When the centers (x0, y0) and (x1, y1) are sufficiently far, the histogram h (l) takes the form of the following equation.

When k = 1, the image shows an object with a sharp border embedded in the background. This object has a dark center and becomes brighter as it goes around. When k = -1, the image represents an object with an ambiguous boundary. This object is brightest at the center and darkens as you move around. Even if you think that a general object is in between these two types of objects, you won't lose generality. Therefore, k can cover most cases with −1 ≦ k ≦ 1, and it is guaranteed that Equation 27 is generally a decreasing function.

  Note that, as can be seen from Equation 34, r is affected by the resolution of the image, i.e., r is proportional to 2 m. For this purpose, a factor of 2 m was introduced in [1.4.1].

ηが０のとき、Ｄ（ｎ）ｆは直前の副写像と無関係に決定され、現在の副写像は弾性的に変形され、過度に歪むことになる。一方、ηが非常に大きな値のとき、Ｄ（ｎ）ｆは直前の副写像によってほぼ完全に決まる。このとき副写像は非常に剛性が高く、画素は同じ場所に射影される。その結果、写像は単位写像になる。ηの値が０から次第に増えるとき、後述のごとくＣ（ｎ）ｆは徐々に減少する。しかしηの値が最適値を越えると、図４に示すとおり、エネルギーは増加し始める。同図のＸ軸はη、Ｙ軸はＣｆである。[1.4.3] Dynamic determination of η The parameter η can be automatically determined by the same method. First, η = 0, and the final map f (n) and energy C (n) f at the finest resolution are calculated. Subsequently, η is increased by a certain value Δη, and the final mapping f (n) and energy C (n) f at the finest resolution are calculated again. This process is continued until the optimum value is obtained. η indicates the rigidity of the mapping. This is because the weight of the following equation.

When η is 0, D (n) f is determined regardless of the immediately preceding submapping, and the current submapping is elastically deformed and excessively distorted. On the other hand, when η is a very large value, D (n) f is almost completely determined by the immediately preceding submapping. At this time, the sub-mapping is very rigid and the pixels are projected to the same location. As a result, the map becomes a unit map. When the value of η increases gradually from 0, C (n) f gradually decreases as described later. However, when the value of η exceeds the optimum value, the energy starts to increase as shown in FIG. In the figure, the X axis is η, and the Y axis is Cf.

  With this method, an optimum value of η that minimizes C (n) f can be obtained. However, as a result of various factors affecting the calculation as compared with the case of λ, C (n) f changes with small fluctuations. In the case of λ, the submapping is only recalculated once every time the input changes by a small amount, but in the case of η, all the submappings are recalculated. For this reason, it cannot be immediately determined whether or not the obtained value of C (n) f is minimum. If a candidate for the minimum value is found, it is necessary to search for the true minimum value by setting a finer section.

における輝度を持つｆ（ｍ，ｓ）が提供される。つまりスーパーサンプリングが行われる。実験では、ｆ（ｍ，ｓ）は整数及び半整数値をとることが許され、

は、

によって与えられた。[1.5] Supersampling When determining the correspondence between pixels, the range of f (m, s) can be expanded to R × R (R is a set of real numbers) in order to increase the degree of freedom. In this case, the luminance of the pixel of the end point image is interpolated, and a non-integer point,

F (m, s) with the brightness at is provided. That is, super sampling is performed. In experiments, f (m, s) is allowed to take integer and half integer values,

Is

Given by.

[1.6] Normalization of luminance of pixels of each image When the start point image and the end point image include very different objects, it is difficult to use the luminance of the original pixel as it is for calculation of mapping. This is because the difference in luminance is large and the energy C (m, s) f relating to luminance becomes too large, making it difficult to perform a correct evaluation.

  For example, consider the case of matching a human face and a cat face. The cat's face is covered with hair and contains very bright and very dark pixels. In this case, the sub-image is first normalized in order to calculate the sub-mapping between the two faces. That is, the brightness of the darkest pixel is set to 0, the brightest is set to 255, and the brightness of the other pixels is obtained by linear interpolation.

[1.7] Implementation Use an inductive method in which the calculation proceeds linearly according to the scan of the starting point image. First, the value of f (m, s) is determined for the top leftmost pixel (i, j) = (0, 0). Next, the value of each f (m, s) (i, j) is determined while increasing i by 1. When the value of i reaches the width of the image, the value of j is increased by 1 and i is returned to 0. Thereafter, f (m, s) (i, j) is determined along with the scan of the starting point image. If pixel correspondence is determined for all points, one mapping f (m, s) is determined.
If the corresponding point qf (i, j) is determined for a certain p (i, j), then the corresponding point qf (i, j + 1) of p (i, j + 1) is determined. At this time, the position of qf (i, j + 1) is limited by the position of qf (i, j) in order to satisfy the bijection condition. Therefore, the priority is higher in this system as the corresponding points are determined first. If (0,0) always has the highest priority, additional deflections are added to the final mapping required. In the base technology, in order to avoid this state, f (m, s) is determined by the following method.

  First, when (s mod 4) is 0, the starting point is (0, 0) and i and j are gradually increased. When (s mod 4) is 1, the right end point of the uppermost row is used as a starting point, and i is decreased and j is increased. When (s mod 4) is 2, the starting point is the rightmost point in the bottom row, and i and j are determined while decreasing. When (s mod 4) is 3, the left end point of the bottom row is used as a starting point, and i is increased and j is decreased. Since the sub-mapping concept, that is, the parameter s does not exist at the nth level with the finest resolution, two directions are calculated continuously assuming that s = 0 and s = 2.

  In an actual implementation, f (m, s) (i, j) satisfying the bijection condition as much as possible from candidates (k, l) is given by giving a penalty to candidates that violate the bijection condition. ) (M = 0, ..., n) was chosen. Candidate energy D (k, l) that violates the third condition is multiplied by φ, while candidates that violate the first or second condition are multiplied by ψ. This time, φ = 2 and ψ = 100000 were used.

ただしここで、

である（ここでベクトルは三次元ベクトルとし、ｚ軸は直交右手座標系において定義される）。もしＷが負であれば、その候補についてはＤ（ｍ，ｓ）（ｋ，ｌ）にψを掛けることによってペナルティを与え、できるかぎり選択しないようにする。In order to check the bijection condition described above, the following test was performed when (k, l) = f (m, s) (i, j) was determined as an actual procedure. That is, it is confirmed whether or not the z component of the outer product of the following equation is 0 or more for each lattice point (k, l) included in the inherited quadrilateral of f (m, s) (i, j).

Where

Where the vector is a three-dimensional vector and the z-axis is defined in an orthogonal right-handed coordinate system. If W is negative, the candidate is penalized by multiplying D (m, s) (k, l) by ψ, and is selected as little as possible.

  FIG. 5A and FIG. 5B show the reason for inspecting this condition. FIG. 5A shows a candidate without a penalty, and FIG. 5B shows a candidate with a penalty. When determining the mapping f (m, s) (i, j + 1) for the adjacent pixel (i, j + 1), if the z component of W is negative, there is no pixel that satisfies the bijection condition on the starting image plane. . This is because q (m, s) (k, l) exceeds the boundary line of the adjacent quadrilateral.

[1.7.1] Submapping Order In implementation, when the resolution level is even, σ (0) = 0, σ (1) = 1, σ (2) = 2, σ (3) = 3, σ (4) = 0 was used, and when odd, σ (0) = 3, σ (1) = 2, σ (2) = 1, σ (3) = 0, and σ (4) = 3 were used. . This effectively shuffled the submap. Note that there are essentially four types of submappings, and s is one of 0 to 3. However, in practice, processing corresponding to s = 4 was performed. The reason will be described later.

つづいてｒ（ｘ，ｙ，ｔ）における画素の輝度が次の式を用いて決定される。

ここでｄｘ及びｄｙはパラメータであり、０から１まで変化する。[1.8] Interpolation calculation After the mapping between the start point image and the end point image is determined, the luminance of the corresponding pixels is interpolated. In the experiment, trilinear interpolation was used. The square p (i, j) p (i + 1, j) p (i, j + 1) p (i + 1, j + 1) in the start image plane is a quadrangle qf (i, j) qf (i + 1, j) qf on the end image plane. Assume that (i, j + 1) qf (i + 1, j + 1) is projected. For simplicity, the distance between images is 1. The pixel r (x, y, t) (0 ≦ x ≦ N−1, 0 ≦ y ≦ M−1) of the intermediate image whose distance from the starting image plane is t (0 ≦ t ≦ 1) is as follows. Is required. First, the position of the pixel r (x, y, t) (where x, y, tεR) is obtained by the following equation.

Subsequently, the luminance of the pixel at r (x, y, t) is determined using the following equation.

Here, dx and dy are parameters and change from 0 to 1.

[1.9] Mapping when imposing constraint conditions We have described the determination of mapping when no constraint conditions exist. However, when a correspondence is defined in advance between specific pixels of the start point image and the end point image, the mapping can be determined using this as a constraint.

  The basic idea is that the starting point image is roughly deformed by a rough mapping in which specific pixels of the starting point image are transferred to specific pixels of the end point image, and then the mapping f is accurately calculated.

  First, a specific pixel of the start point image is projected onto a specific pixel of the end point image, and a rough mapping for projecting other pixels of the start point image to an appropriate position is determined. That is, a pixel that is close to a specific pixel is a mapping that is projected near the place where the specific pixel is projected. Here, a rough mapping at the m-th level is described as F (m).

を特定するとき、以下の値を決める。

始点画像の他の画素の変位量は、ｐ（ｉｈ，ｊｈ）（ｈ＝０，…，ｎｓ−１）の変位に重み付けをして求められる平均である。すなわち画素ｐ（ｉ，ｊ）は、終点画像の以下の画素に射影される。

ただしここで、

とする。The rough map F is determined as follows. First, a mapping is specified for several pixels. Ns pixels for the starting image,

When specifying, the following values are determined.

The displacement amount of the other pixels of the starting point image is an average obtained by weighting the displacement of p (ih, jh) (h = 0,..., Ns−1). That is, the pixel p (i, j) is projected onto the following pixels of the end point image.

Where

And

である。ただし、

であり、κ，ρ≧０とする。最後に、前述の写像の自動計算プロセスにより、ｆを完全に決定する。Subsequently, the energy D (m, s) (i, j) of the map f is changed so that the candidate map f near F (m) has less energy. To be precise, D (m, s) (i, j) is

It is. However,

And κ, ρ ≧ 0. Finally, f is completely determined by the automatic map calculation process described above.

以内であるとき、Ｅ２（ｍ，ｓ）（ｉ，ｊ）が０になることに注意すべきである。そのように定義した理由は、各ｆ（ｍ，ｓ）（i,j）がＦ（ｍ）（i,j）に十分近い限り、終点画像において適切な位置に落ち着くよう、その値を自動的に決めたいためである。この理由により、正確な対応関係を詳細に特定する必要がなく、始点画像は終点画像にマッチするように自動的にマッピングされる。
［２］具体的な処理手順
［１］の各要素技術による処理の流れを説明する。
図６は前提技術の全体手順を示すフローチャートである。同図のごとく、まず多重解像度特異点フィルタを用いた処理を行い（Ｓ１）、つづいて始点画像と終点画像のマッチングをとる（Ｓ２）。ただし、Ｓ２は必須ではなく、Ｓ１で得られた画像の特徴をもとに画像認識などの処理を行ってもよい。Here, when f (m, s) (i, j) is sufficiently close to F (m) (i, j), that is, their distance is

Note that E2 (m, s) (i, j) is 0 when within. The reason for this definition is that as long as each f (m, s) (i, j) is sufficiently close to F (m) (i, j), its value is automatically set so as to settle at an appropriate position in the end point image. Because we want to decide. For this reason, it is not necessary to specify an exact correspondence relationship in detail, and the start point image is automatically mapped so as to match the end point image.
[2] Specific Processing Procedure A processing flow according to each elemental technique of [1] will be described.
FIG. 6 is a flowchart showing the overall procedure of the base technology. As shown in the figure, first, processing using a multi-resolution singularity filter is performed (S1), and then the start point image and the end point image are matched (S2). However, S2 is not essential, and processing such as image recognition may be performed based on the characteristics of the image obtained in S1.

  FIG. 7 is a flowchart showing details of S1 of FIG. Here, it is assumed that the start point image and the end point image are matched in S2. Therefore, first, the starting point image is hierarchized by a singular point filter (S10), and a series of starting point hierarchical images are obtained. Subsequently, the end point images are hierarchized by the same method (S11), and a series of end point layer images are obtained. However, the order of S10 and S11 is arbitrary, and the start point hierarchical image and the end point hierarchical image can be generated in parallel.

  FIG. 8 is a flowchart showing details of S10 in FIG. The size of the original starting point image is 2n × 2n. Since the start point hierarchical image is created in order from the finer resolution, the parameter m indicating the resolution level to be processed is set to n (S100). Subsequently, a singular point is detected from the m-th level images p (m, 0), p (m, 1), p (m, 2), and p (m, 3) using a singular point filter (S101). The m-1st level images p (m-1, 0), p (m-1, 1), p (m-1, 2), and p (m-1, 3) are generated (S102). Here, since m = n, p (m, 0) = p (m, 1) = p (m, 2) = p (m, 3) = p (n), and 4 from one start point image. A type of sub-image is generated.

  FIG. 9 shows a correspondence relationship between a part of the mth level image and a part of the m−1th level image. The numerical values in the figure indicate the luminance of each pixel. P (m, s) in the figure symbolizes four images of p (m, 0) to p (m, 3). When generating p (m-1, 0), p (m-1, 0) m, s) is considered to be p (m, 0). According to the rule shown in [1.2], p (m-1, 0) is, for example, “3” of the four pixels included in the block in which the luminance is entered in FIG. ) Is “8”, p (m−1,2) is “6”, p (m−1,3) is “10”, and this block is replaced with one acquired pixel. Accordingly, the size of the sub-image at the (m−1) th level is 2m−1 × 2m−1.

  Next, m is decremented (S103 in FIG. 8), it is confirmed that m is not negative (S104), and the process returns to S101 and a sub-image with a coarse resolution is generated. As a result of this iterative process, S10 ends when m = 0, that is, when the 0th level sub-image is generated. The size of the 0th level sub-image is 1 × 1.

  FIG. 10 illustrates the start point hierarchical image generated in S10 for n = 3. Only the first starting point image is common to the four sequences, and thereafter, sub-images are generated independently according to the types of singular points. Note that the processing in FIG. 8 is common to S11 in FIG. 7, and an end point hierarchical image is also generated through the same procedure. Thus, the process by S1 in FIG. 6 is completed.

となる。ただし、総和はｉ、ｊについてそれぞれ０、１…、２ｍ−１で計算する。以上でマッチング評価の準備が整う。In the base technology, preparation for matching evaluation is made in order to proceed to S2 in FIG. FIG. 11 shows the procedure. As shown in the figure, first, a plurality of evaluation formulas are set (S30). The energy C (m, s) f related to the pixel introduced in [1.3.2.1] and the energy D (m, s) f related to the smoothness of the map introduced in [1.3.2.2] is there. Next, a comprehensive evaluation formula is established by integrating these evaluation formulas (S31). That is the total energy λC (m, s) f + D (m, s) f introduced in [1.3.2.3], and using η introduced in [1.3.2.2],

It becomes. However, the sum is calculated with 0, 1,..., 2m−1 for i and j, respectively. The matching evaluation is now complete.

  FIG. 12 is a flowchart showing details of S2 of FIG. As described in [1], matching between the start point hierarchical image and the end point hierarchical image is performed between images having the same resolution level. In order to obtain good global matching between images, matching is calculated in order from the level with the coarser resolution. Since the start and end layer images are generated using the singularity filter, the position and brightness of the singularity are clearly preserved even at a coarse level of resolution, and the results of global matching are compared to the conventional case. It will be very good.

As shown in FIG. 12, first, the coefficient parameter η is set to 0 and the level parameter m is set to 0 (S20). Subsequently, a match is calculated between each of the four sub-images at the m-th level in the start point hierarchical image and the four sub-images at the m-th level in the end point hierarchical image, satisfying the bijection condition, and energy. 4 types of sub-mappings f (m, s) (s = 0, 1, 2, 3) are obtained (S21). The bijection condition is checked using the inherited quadrilateral described in [1.3.3]. At this time, as shown in Expressions 17 and 18, since the submapping at the m-th level is constrained to those at the (m-1) -th level, matching at a level with a coarser resolution is sequentially used. This is a vertical reference between different levels. Although m = 0 and there is no coarser level, this exceptional processing will be described later with reference to FIG.
On the other hand, horizontal reference within the same level is also performed. As shown in Equation 1 of [1.3.3], f (m, 3) is f (m, 2), f (m, 2) is f (m, 1), and f (m, 1) is It is determined to be similar to f (m, 0). The reason is that even if the types of singular points are different, it is unnatural that the submappings are completely different as long as they are originally included in the same start point image and end point image. As can be seen from Equation 20, the closer the sub-maps, the smaller the energy and the better the matching.

  For f (m, 0) to be determined first, there is no submapping that can be referred to at the same level, so one coarse level is referenced as shown in Equation 19. However, in the experiment, after obtaining up to f (m, 3), the procedure was taken to update f (m, 0) once using this as a constraint. This is equivalent to substituting s = 4 into Equation 20 to make f (m, 4) a new f (m, 0). This is to avoid the tendency that the relation between f (m, 0) and f (m, 3) becomes too low, and this measure has improved the experimental results. In addition to this measure, the sub-map shuffle shown in [1.7.1] was also performed in the experiment. This is also intended to keep the degree of association between the submaps originally determined for each type of singularity closely. In addition, as described in [1.7], the position of the start point is changed according to the value of s in order to avoid deflection depending on the start point of processing.

  FIG. 13 is a diagram showing how the submapping is determined at the 0th level. At the 0th level, since each sub-image is composed of only one pixel, all four sub-maps f (0, s) are automatically determined as unit maps. FIG. 14 is a diagram showing how the submapping is determined at the first level. At the first level, each sub-image is composed of 4 pixels. In the figure, these four pixels are indicated by solid lines. Now, when searching for a corresponding point of the point x of p (1, s) in q (1, s), the following procedure is taken.

1. The upper left point a, the upper right point b, the lower left point c, and the lower right point d of the point x are obtained at the first level resolution.
2. The pixel to which the points a to d belong at one coarse level, that is, the 0th level is searched. In the case of FIG. 14, the points a to d belong to the pixels A to D, respectively. However, the pixels A to C are virtual pixels that do not originally exist.
3. The corresponding points A ′ to D ′ of the pixels A to D already obtained at the 0th level are plotted in q (1, s). Pixels A ′ to C ′ are virtual pixels, and are assumed to be in the same positions as the pixels A to C, respectively.
4). The corresponding point a ′ of the point a in the pixel A is considered to be in the pixel A ′, and the point a ′ is plotted. At this time, it is assumed that the position occupied by the point a in the pixel A (in this case, the lower right) is the same as the position occupied by the point a ′ in the pixel A ′.
Corresponding points b ′ to d ′ are plotted in the same manner as in 5.4, and an inherited quadrilateral is created with the points a ′ to d ′.
6). The corresponding point x ′ of the point x is searched so that the energy is minimized in the inherited quadrilateral. The candidate for the corresponding point x ′ may be limited to, for example, a pixel whose center is included in the inherited quadrilateral. In the case of FIG. 14, all four pixels are candidates.

  This is the procedure for determining the corresponding point of point x. Similar processing is performed for all other points to determine the submapping. At the level higher than the second level, it is considered that the shape of the inherited quadrilateral gradually collapses. Therefore, as shown in FIG.

  When four sub-mappings at a certain mth level are determined in this way, m is incremented (S22 in FIG. 12), it is confirmed that m does not exceed n (S23), and the process returns to S21. Thereafter, every time the process returns to S21, a sub-mapping of a finer resolution level is gradually obtained, and when returning to S21 at the end, an n-th level map f (n) is determined. Since this mapping is determined with respect to η = 0, it is written as f (n) (η = 0).

  Next, η is shifted by Δη and m is cleared to zero in order to obtain a mapping for different η (S24). After confirming that the new η does not exceed the predetermined search truncation value ηmax (S25), the process returns to S21, and the mapping f (n) (η = Δη) is obtained for the current η. This process is repeated, and f (n) (η = iΔη) (i = 0, 1,...) Is obtained in S21. When η exceeds ηmax, the process proceeds to S26, an optimum η = ηopt is determined by a method described later, and f (n) (η = ηopt) is finally set as a mapping f (n).

  FIG. 15 is a flowchart showing details of S21 of FIG. According to this flowchart, the submapping at the m-th level is determined for a certain η. When determining the submapping, the base technology determines the optimum λ for each submapping independently.

  As shown in the figure, s and λ are first cleared to zero (S210). Next, a submapping f (m, s) that minimizes energy is obtained for λ at that time (and implicitly for η) (S211), and this is written as f (m, s) (λ = 0). In order to obtain a mapping with respect to different λ, λ is shifted by Δλ, it is confirmed that the new λ does not exceed the predetermined search truncation value λmax (S213), the process returns to S211 and f (m, s) (λ = iΔλ) (i = 0, 1,...) is obtained. When λ exceeds λmax, the process proceeds to S214, where an optimal λ = λopt is determined, and f (m, s) (λ = λopt) is finally set as a mapping f (m, s) (S214).

  Next, to obtain another submapping at the same level, λ is cleared to zero and s is incremented (S215). It is confirmed that s does not exceed 4 (S216), and the process returns to S211. When s = 4, f (m, 0) is updated using f (m, 3) as described above, and the submapping at that level is finished.

  FIG. 16 shows the behavior of energy C (m, s) f corresponding to f (m, s) (λ = iΔλ) (i = 0, 1,...) Obtained while changing λ for certain m and s. FIG. As described in [1.4], when λ increases, C (m, s) f usually decreases. However, when λ exceeds the optimum value, C (m, s) f starts to increase. Therefore, in the base technology, λ when C (m, s) f takes a minimum value is determined as λopt. Even if C (m, s) f becomes smaller again in the range of λ> λopt as shown in the figure, the mapping is already broken at that point and it does not make sense, so if we pay attention to the first minimum point, Good. λopt is determined independently for each submapping, and finally f (n) is also determined.

  On the other hand, FIG. 17 is a diagram showing the behavior of energy C (n) f corresponding to f (n) (η = iΔη) (i = 0, 1,...) Obtained while changing η. Again, when η increases, C (n) f usually decreases, but when η exceeds the optimum value, C (n) f starts to increase. Therefore, η when C (n) f takes the minimum value is determined as ηopt. FIG. 17 may be considered as an enlarged view of the vicinity of zero on the horizontal axis of FIG. If ηopt is determined, f (n) can be finally determined.

  As described above, according to the base technology, various advantages can be obtained. First, since it is not necessary to detect an edge, it is possible to solve the problem of the conventional technology of the edge detection type. In addition, a priori knowledge about the object included in the image is unnecessary, and automatic detection of corresponding points is realized. The singularity filter can maintain the luminance and position of the singularity even at a coarse resolution level, and is extremely advantageous for object recognition, feature extraction, and image matching. As a result, it is possible to construct an image processing system that greatly reduces manual work.

It should be noted that the following modification technique is also conceivable for this prerequisite technique.
(1) In the base technology, the parameters are automatically determined when matching between the start layer image and the end layer image. However, this method performs matching between two normal images, not between layer images. If available in general.

  For example, between two images, the energy E0 related to the difference in pixel brightness and the energy E1 related to the positional deviation of the pixels are used as evaluation expressions, and the linear sum Etot = αE0 + E1 is used as a comprehensive evaluation expression. Focusing on the vicinity of the extreme value of this comprehensive evaluation formula, α is automatically determined. That is, a mapping that minimizes Etot for various α is obtained. Among these maps, α when E1 takes a minimum value with respect to α is determined as an optimum parameter. The mapping corresponding to that parameter is finally regarded as the optimal matching between both images.

  In addition to this, there are various methods for setting the evaluation formula. For example, a value that takes a larger value as the evaluation result is better, such as 1 / E1 and 1 / E2, may be adopted. The comprehensive evaluation formula is not necessarily a linear sum, and an n-th power sum (n = 2, 1/2, −1, −2 etc.), a polynomial, an arbitrary function, or the like may be selected as appropriate.

The parameter may be either α, two cases of η and λ as in the base technology, or more than that. If the parameter is 3 or more, change it one by one.
(2) In the base technology, after determining the mapping so that the value of the comprehensive evaluation formula is minimized, a point where C (m, s) f, which is one evaluation formula constituting the comprehensive evaluation formula, is minimized is detected. Parameters were determined. However, instead of such a two-stage process, it is effective to determine parameters so that the minimum value of the comprehensive evaluation formula is minimized in some situations. In that case, for example, αE0 + βE1 may be set as a comprehensive evaluation formula, and a constraint condition of α + β = 1 may be provided to treat each evaluation formula equally. This is because the essence of automatic parameter determination is that the parameter is determined so that the energy is minimized.
(3) In the base technology, four types of sub-images relating to four types of singular points are generated at each resolution level. However, of course, one, two, and three of the four types may be selectively used. For example, if there is only one bright point in the image, a corresponding effect should be obtained even if a hierarchical image is generated using only f (m, 3) related to the maximum point. In this case, since different submappings at the same level are not required, there is an effect of reducing the amount of calculation regarding s.
(4) In the base technology, when the level is advanced by one by the singularity filter, the pixel becomes 1/4. For example, a configuration in which 3 × 3 is one block and a singular point is searched for is possible. In this case, the pixel becomes 1/9 when the level advances by one.
(5) If the start point image and the end point image are in color, they are first converted to a black and white image and a mapping is calculated. The starting point color image is converted using the mapping obtained as a result. As another method, a submapping may be calculated for each component of RGB.
[3] Improvements of Premise Technology Based on the above premise technologies, some improvements have been made to improve matching accuracy. Here are the improvements.

ここで画素ａにおけるＹ（輝度）をＹ(ａ)として、次のような記号を定義する。

上の定義を用いて以下のような５つのフィルタを用意する。

このうち上から４つのフィルタは改良前の前提技術におけるフィルタとほぼ同じで、輝度の特異点を色情報も残しながら保存する。最後のフィルタは色の彩度の特異点をこちらも色情報を残しながら保存する。[3.1] Singularity filter and color image taking color information into consideration In order to effectively use the color information of the sub-image, the singularity filter was changed as follows. First, as the color space, HIS, which is said to best match human intuition, was used, and the formula that converts colors into luminance was chosen to be the closest to the human eye sensitivity.

Here, Y (luminance) in the pixel a is defined as Y (a), and the following symbols are defined.

The following five filters are prepared using the above definition.

Of these, the four filters from the top are almost the same as the filters in the base technology before the improvement, and the singular points of luminance are preserved while retaining the color information. The last filter saves the singularity of the color saturation while leaving the color information.

［３．２］エッジ画像およびその副画像
輝度微分（エッジ）の情報をマッチングに利用するため、一次微分エッジ検出フィルタを用いる。このフィルタはあるオペレータＨとの畳み込み積分で実現できる。

ここでＨは演算スピードなども考慮し、以下のようなオペレータを用いた。

次にこの画像を多重解像度化する。フィルタにより０を中心とした輝度をもつ画像が生成されるため、次のような平均値画像が副画像としては最も適切である。

式５９の画像は後述するForward Stage、すなわち初回副写像導出ステージの計算の際、エネルギー関数に用いられる。With these filters, five types of sub-images (sub-images) are generated for each level. Note that the highest level sub-image matches the original image.

[3.2] Edge image and its sub-image In order to use information on luminance differentiation (edge) for matching, a primary differential edge detection filter is used. This filter can be realized by convolution integration with an operator H.

Here, in consideration of calculation speed and the like, H used the following operator.

Next, this image is converted to multi-resolution. Since an image having luminance centered at 0 is generated by the filter, the following average image is most suitable as a sub-image.

The image of Expression 59 is used as an energy function when calculating a forward stage, that is, a first submapping derivation stage described later.

この値は常に正であるため、多重解像度化には最大値フィルタを用いる。

式６１の画像は後述するForward Stageの計算の際、計算する順序を決定するのに用いられる。The size of the edge, that is, the absolute value is also necessary for the calculation.

Since this value is always positive, a maximum value filter is used for multiresolution.

The image of Formula 61 is used to determine the order of calculation when calculating the Forward Stage described later.

[3.3] Calculation processing procedure The calculation is performed in order from the sub-image with the coarsest resolution. Since there are five sub-images, the calculation is performed multiple times at each level of resolution. This is called a turn, and the maximum number of calculations is represented by t. Each turn consists of two energy minimization calculations, the Forward Stage and the Refinement Stage, which is a submapping recalculation stage. FIG. 18 is a flowchart according to the improvement in the calculation for determining the submapping at the m-th level.

  As shown in the figure, s is cleared to zero (S40). Next, in the forward stage (S41), a map f (m, s) from the start point image p to the end point image q is obtained by energy minimization. Here, the energy to be minimized is a linear sum of energy C by the corresponding pixel value and energy D by the smoothness of the mapping.

エネルギーＤは前記改良前の前提技術と同じものを用いる。ただし前記改良前の前提技術において、写像の滑らかさを保証するエネルギーＥ１を導出する際、隣接する画素のみを考慮していたが、周囲の何画素を考慮するかをパラメータｄで指定できるように改良した。

次のRefinement Stageに備えて、このステージでは終点画像ｑから始点画像ｐへの写像ｇ（ｍ，ｓ）も同様に計算する。The energy C is composed of energy CI (equivalent to energy C in the pre-improvement premise technology), luminance CC, energy CC due to hue and saturation, and energy CE due to luminance differential (edge) difference. It is expressed as follows.

The energy D is the same as the base technology before the improvement. However, in the base technology before the improvement, when deriving the energy E1 that guarantees the smoothness of the mapping, only the adjacent pixels are taken into consideration, but the number of surrounding pixels can be specified by the parameter d. Improved.

In preparation for the next Refinement Stage, the mapping g (m, s) from the end point image q to the start point image p is similarly calculated in this stage.

対称性を損なわないように、終点画像ｑから始点画像ｐへの写像ｇ’（ｍ，ｓ）も同様の方法で求めておく。In the Refinement Stage (S42), a more appropriate map f ′ (m, s) is obtained based on the bidirectional maps f (m, s) and g (m, s) obtained in the Forward Stage. Here, energy minimization calculation is performed for the newly defined energy M. The energy M is composed of the degree of matching M0 with the mapping g from the end point image to the starting point image, and the difference M1 between the original mapping.

The mapping g ′ (m, s) from the end point image q to the start point image p is also obtained in the same way so as not to impair the symmetry.

［３．４］写像の計算順序
写像の滑らかさを表すエネルギーＥ１を計算する際、周囲の点の写像を用いるため、それらの点がすでに計算されているかどうかがエネルギーに影響を与える。すなわち、どの点から順番に計算するかによって、全体の写像の精度が大きく変化する。そこでエッジの絶対値画像を用いる。エッジの部分は情報量を多く含むため、エッジの絶対値が大きいところから先に写像計算を行う。このことによって、特に二値画像のような画像に対して非常に精度の高い写像を求めることができるようになった。Thereafter, s is incremented (S43), it is confirmed that s does not exceed t (S44), and the process proceeds to the Forward Stage (S41) of the next turn. At that time, the energy minimization calculation is performed by replacing E0 as follows.

[3.4] Map Calculation Order Since the map of surrounding points is used when calculating the energy E1 representing the smoothness of the map, whether or not those points have already been calculated affects the energy. That is, the accuracy of the overall mapping varies greatly depending on which point is calculated in order. Therefore, an edge absolute value image is used. Since the edge portion contains a large amount of information, the mapping calculation is performed first from the point where the absolute value of the edge is large. This makes it possible to obtain a highly accurate mapping, particularly for images such as binary images.

  In the base technology, matching between two key frames is performed to generate corresponding position information (hereinafter also referred to as “corresponding point information”), and an intermediate frame is generated based on the position information. The key frame is an image to be matched, and is expressed as a start point image and an end point image in the base technology. This technology can be used to compress moving images, and in reality, both image quality and compression ratio exceeding MPEG have been confirmed in experiments.

(First embodiment)
Hereinafter, the encoding technique and decoding technique according to the first embodiment will be described in order.
[1] Image Coding Technology First, the image coding technology according to the first embodiment will be described with reference to FIG. FIG. 19 shows the flow of the present image encoding technique, and at the same time, shows the configuration of an image encoding apparatus to be described later. In the figure, each element described as a functional block for performing various processes can be configured with a CPU, a memory, and other LSIs in terms of hardware, and loaded into the memory in terms of software. Realized by programs. Therefore, it is understood by those skilled in the art that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof, and is not limited to any one. Further, in the figure, when the same block appears in a plurality of places, it does not necessarily mean that there are a plurality of the same blocks, but may mean that one block is used a plurality of times.
(1) A sequence of image frames to be processed (hereinafter simply referred to as “target image frames”) is expressed as F0, F1,..., Fn−1, Fn (n is an integer of 2 or more) in descending or ascending order. To do. Corresponding point information indicating the positional relationship of the points corresponding to each other between the image frames Fi and Fj (i, j = 0, 1,..., N) is denoted as Mi-j. The points on the image frame Fi are collectively written as pi. The target image frames may be equally spaced in time or not. At this time, the encoding technique performs the following steps.

Step a
Matching is calculated between the image frames F0 and Fn to generate corresponding point information M0-n. Matching is a process for specifying a region or a point that corresponds between image frames. When the base technology is used, matching is performed on a pixel basis, and block matching is used for matching similar to MPEG. When the point p0 on the image frame F0 and the point pn on the image frame Fn correspond to each other, the simplest description example of the corresponding point information M0-n is “p0 → pn”. This is actually described by the coordinates in the image (hereinafter simply referred to as “coordinates”). Specifically, when the point pi is composed of pi0, pi1,... And they correspond to each other, a description example of the corresponding point information M0-n is as follows.
“P00 → pn0 / p01 → pn1 / p02 → pn2 //.”

Step b
A path for moving the point p0 on the image frame F0 to the corresponding point pn on the image frame Fn by the corresponding point information M0-n is divided into n, and the point p1 corresponding to the point p0 on the image frame F1 and the image frame F2 To calculate a point p2 corresponding to the point p0,..., A point pn corresponding to p0 on the image frame Fn. If the target image frames are equally spaced in time, “n division” is divided into n equal parts. If not, division by a division ratio according to the time ratio between image frames is performed. For example, when the coordinates of p0 are (x0, y0), that of pn is (xn, yn), and the target image frames are equally spaced, the coordinates of pi are generalized as follows.
((Xn−x0) · i / n + x0, (yn−y0) · i / n + y0)
This is a coordinate calculation by so-called interpolation. This description is an example in which the corresponding points of F0 and Fn are linearly interpolated, but there is also interpolation by a curve as will be described later.

Step c
By executing step b for a predetermined number of points on the image frame F0, using a set of points p1 corresponding to the predetermined number of points, using a set of virtual image frames F1 ′ and p2 A virtual image frame Fn ′ is generated using a set of virtual image frames F2 ′,. An example of the “predetermined number of points” is all the pixels constituting the image frame. However, in this case, the amount of calculation also increases, and for example, one pixel may be extracted in several pixels in the x and y directions of the image frame. This is the same as dividing an image frame into meshes and extracting only the pixels corresponding to the mesh grid points. For example, if one pixel is extracted for every five pixels in the x and y directions, the “predetermined number of points” is 1/25 of the total number of pixels.

  If the “predetermined number” is not the total number of pixels, the corresponding points for which the corresponding points were not calculated (interpretally called non-grid points) are interpolated based on the calculated corresponding points (tentatively called grid points). Is calculated by For example, a non-grid point on the image frame F0 is between two grid points on the same F0, the former position vector is p, the latter two are q, r, and p = (1-α) q + αr. The point p ′ corresponding to the point p on the image frame Fn can be expressed as p ′ = (1−α) q using the points q ′ and r ′ corresponding to q and r on the same Fn, respectively. It can be written as “+ αr”. A method of generalizing this and describing a non-grid point with three grid points is known as bilinear interpolation. This can be used.

Step d
A set S1 of a virtual image frame F1 ′ and a real image frame F1, a set S2 of a virtual image frame F2 ′ and a real image frame F2,..., A virtual image frame Fn ′ and a real image frame For each set of Fn sets Sn, the presence or absence of a set Sk (k = 1,..., N) having a large difference between image frames included in the set is determined based on a predetermined determination criterion.
The “predetermined determination criterion” may be “large” when the difference is simply compared with a predetermined threshold value and exceeded. That is, you may pay attention to the difference itself. Here, the threshold value may be determined by experiment, and the same can be said for other parameters.

  As another “determination criterion”, an energy value calculated by the base technology, that is, a physical quantity that suggests the magnitude of the difference may be used. The energy increases as the positions of the corresponding points are farther away, and the energy increases as the pixel value increases. Therefore, in general, the greater the energy, the higher the possibility that the response is not accurately captured. If the correspondence is not accurate, the difference tends to increase. Therefore, if the energy value between image frames is larger than a predetermined threshold value, it may be determined that the difference is large.

Steps e and f
When there is a set Sk having a large difference, at least matching between the image frame Fh (h = 0, 1,..., K−1) and the image frame Fk ′ is generated to generate corresponding point information Mh-k. To do. Subsequently, the corresponding point information M0-n is corrected using the information of the corresponding point information Mh-k (hereinafter, the corresponding point information M0-n before correction is simply referred to as “original M0-n” or “M0-n”. The corrected corresponding point information M0-n is referred to as “corrected M0-n” or “M0-n ′”). As an example, if h = 0, M0-k is obtained, and a point pk corresponding to the point p0 is obtained. The correction M0-n is, for example, “p0 → pk → pn”, which is a representation in a polygonal line format. Thereby, since the point p0 reaches pn via pk, the difference of Sk becomes smaller than the linear form expression by the original M0-n. The correction M0-n may be a curved line instead of a polygonal line. In that case, the trajectory may be described by, for example, a spline curve so that “p0 → pk → pn” or its approximation is realized.

Step g
Coded data in a format including at least the image frame F0 and the modified M0-n is output. If there is an image frame F0 and a modification M0-n, the point p0 on the image frame F0 can be reached to pn via pk, and therefore it is established as encoded data. The encoded data may include image frame Fn data. In that case, not only p0 moves but also the pixel value can be changed by interpolation. If the pixel value of the point p0 is Vp0 and that of the point pn is Vpn, the pixel value Vpi of the point pi can be interpolated as follows. Non-grid points may be obtained by bilinear interpolation as in the case of coordinates.
Vpi = (Vpn−Vp0) · i / n + Vp0

  However, since it can be considered that the points p0 and pn are detected as corresponding points because the pixel values are also close to each other, it is not essential to describe the change in the pixel value. Therefore, only F0 of the image frames F0 and Fn is established as encoded data.

  According to the above method, relatively high image quality can be realized with relatively small data. This is because it is sufficient that there is at least one piece of data as an image frame, and the data amount of the correction M0-n is greatly reduced by appropriately reducing the “predetermined number of points”. Therefore, since the total amount of data is small, Sk having a large difference is dealt with by correcting the original M0-n, so that the image quality is greatly improved. This fact has been confirmed by experiments.

(2) In step d, the magnitude of the difference between image frames is determined in units of predetermined areas. The region is obtained by simply dividing an image frame into meshes, for example. The size of the area may be selected from a combination of image quality and data amount by experiment.
When determining in units of regions in step d, when the virtual image frame Fk ′ and the actual image frame Fk correspond to each other, that is, when the difference between the same regions in position is larger than a predetermined threshold value, Sk is determined as a “group with a large difference”. In this case, the sum of the differences over the entire image frame between the virtual image frame Fk ′ and the actual image frame Fk is not necessarily large, and it is not necessary. It is only necessary to find an area with a large difference. Here, instead of paying attention to the difference itself, or in addition thereto, a total energy may be calculated for each region and compared with a threshold value. Hereinafter, in the embodiment, comparison is made for each region, not for the entire image frame.

  20A to 20D are diagrams illustrating examples of target image frames. The comparison for each region is particularly effective, as shown in FIGS. 20A to 20C, in which the target image frame shows “ball bounce”, and the image frame F0 in FIG. The image frame Fn shown in FIG. 20C corresponds to the case where the image frame Fk shown in FIG. Even if F1, F2,... Are generated by interpolation using the original M0-n, the ball only moves linearly from the position before bouncing to the position after bouncing, which is unnatural. 20A to 20C, the actual trajectory of the ball is shown by a solid line, and the trajectory of the ball generated by interpolation using the original M0-n is shown by a broken line. If the difference is calculated for each region, Sk is obtained by adjusting the threshold value, and the group before and after that can be detected. For example, in the examples of FIGS. 20A to 20C, in the image frame Fk, the difference between the actual ball Bk and the region including the ball position Bk ′ generated by interpolation exceeds the threshold, and Sk is Detected. If they can be detected, they can be reflected in the correction M0-n, so that the state of the ball bouncing can be expressed more accurately. FIG. 20D shows the trajectory of the ball reproduced based on the correction M0-n.

  Note that a plurality of areas with large differences may be detected in the same Sk. For example, it is assumed that the point p00 on the image frame F0 is included in the region A having a large difference and the point p01 is included in another region B having a large difference. Also in this case, the correction M0-n has a format of “p00 → pn0 / p01 → pn1 /...”, And eventually, even if a plurality of areas are detected, the format of the correction M0-n itself is not affected. For the same reason, a plurality of regions may be detected in different sets.

(3) Step e may include the following sub-steps.
e1) When a set Sk having a large difference exists, corresponding point information M0-1 between the image frames F0 and F1, M1-2 between F1 and F2,..., M (k− between Fk−1 and Fk) 1) Find each -k.
e2) M0-1, M1-2,..., M (k-1) -k are integrated to generate M0-k.
This has already been described.

(4) In step e2, p1 corresponding to the point p0 is obtained from the corresponding point information M0-1, p2 corresponding to the point p1 is obtained from the corresponding point information M1-2,..., Corresponding point information M (k− 1) Obtain pk corresponding to the point pk-1 by -k, and specify the points corresponding to p0 in order of p1, p2,..., Pk, and finally specify pk corresponding to p0. M0-k may be generated. This has already been described.

(5) Step f may use the information of corresponding point information M0-k, and generate correction M0-n in a format indicating a locus where point p0 reaches pn via pk. This has already been described.

(6) If it is determined in step d that a set Sk having a large difference exists, step g outputs encoded data in a format including information on the difference in the set Sk in addition to the image frame F0 and the modified M0-n. May be. For example, when the difference in the region A of the set Sk is large, it is conceivable that complete image quality cannot be obtained only by changing the original M0-n to the modified M0-n. In that case, first, the difference may be reduced by a considerable amount by correcting the corresponding point information M0-n, and the remaining difference may be further described in the encoded data as the difference in the area A. In this case, the format of the encoded data is, for example, as follows.

FIG. 21 is a diagram illustrating a format of encoded data in the image encoding technique according to the first embodiment. The encoded data D1 includes the image frame (i), the modified corresponding point information (ii), the presence / absence bit (iii), the difference information (iv), the value of k (v), the position / And a shape method (vi). The contents of each data are as follows.
i) F0 data or F0 + Fn data
ii) Modified M0-n (form of broken line, curve, etc.)
iii) “Presence / absence bit” indicating the presence / absence of difference information
iv) Difference information (image data format)
v) Value of k (in the case of multiple k1, k2,...)
vi) Position / shape information of the area (in the case of multiple areas, A1 (k1), A2 (k2),...)

Here, the difference information is “data relating to the area A between the virtual Fk ′ and the actual Fk”, and takes the form of the image data of the area A. The difference information is valid only when the presence / absence bit affirms the presence of the difference information. When the presence / absence bit is negative, the information below the difference information is ignored in the later-described image decoding.
Since the difference information is in the form of image data, it is desirable that the difference information is compressed by a known compression method and then stored in the encoded data. The difference information has no meaning as an image, and it is easy to make a clear statistical bias around zero, which is advantageous in that a relatively high compression rate can be realized.
The value of k and the position / shape information of the region indicate which part of which set Sk the difference information relates to. In the decoding apparatus, the difference is appropriately added based on the value of k and the position / shape information of the region.

(7) The difference information in the set Sk is included in the encoded data only for the region where the difference is large in the image frame. This has already been described.

(8) Along with the difference, at least the value of k and the position information of the area A are included in the encoded data. This has already been described.

(9) The difference is included in the encoded data after being subjected to compression processing. This has already been described.
(10) An aspect of the image encoding device includes the following configuration. The processing operation itself of each configuration has already been described.
・ Matching processing unit 20
A matching is calculated between F0 and Fn to generate M0-n.
Intermediate frame generator 22
A path for moving the point p0 on F0 to the corresponding point pn on Fn by M0-n is divided into n points, the point p1 corresponding to p0 on F1, the point p2 corresponding to p0 on F2,. By executing the process of calculating the point pn corresponding to p0 on Fn with respect to a predetermined number of points on F0, a virtual F1 ′, point using the set of points p1 corresponding to the predetermined number of points Virtual F2 ′ is generated using the set of p2 and virtual Fn ′ is generated using the set of points pn.

・ Determining unit 24
Virtual F1 ′ and real F1 set S1, virtual F2 and real F2 set S2,..., Virtual Fn and real Fn set Sn included in the set The presence / absence of a set Sk (k = 1,..., N) having a large difference between image frames to be determined is determined based on a predetermined determination criterion.
Based on the above configuration, when there is a set Sk having a large difference, the matching processing unit calculates matching between at least Fh (h = 0, 1,..., K−1) and Fk to obtain Mh. -K is generated, and M0-n is corrected using the information of Mh-k. The apparatus further includes an output unit 40 that outputs encoded data in a format including at least F0 and modified M0-n.

(11) Another aspect of the image encoding method performs the following processing.
Process 1: Matching calculation is executed between the two end image frames of an image group including three or more image frames. Examples of both-end image frames are F0 and Fn described above.
Process 2: Based on the corresponding point information between the two-end image frames obtained as a result of the matching calculation, an intermediate image frame sandwiched between the two-end image frames is virtually generated by interpolation. In this example, F1 ′, F2 ′,... Are generated by the interpolation described above.

Process 3: For any region on the image, whether or not any of the virtually generated intermediate image frames has a difference greater than the allowable value from the actual intermediate image frame is determined based on a predetermined criterion. Judge with. That is, a comparison is performed in units of regions between two image frames. This example has already been described as set Sk. Here, it is referred to as “difference greater than or equal to the allowable value”, and if it is less than the allowable value, the process 4 can be skipped.
Process 4: Encoded data including at least one of both-end image frames and corresponding point information is generated. When it is determined that there is an area having a difference greater than or equal to the allowable value, difference information regarding the area is also generated. This has already been described.

(12) Another aspect of the image encoding device includes the following configuration. The processing by each configuration has already been described.
・ Matching processing unit 20
Matching calculation is performed between the two end image frames of the image group including three or more image frames.
Intermediate frame generator 22
Based on the corresponding point information between the two-end image frames obtained as a result of the matching calculation, an intermediate image frame sandwiched between the two-end image frames is virtually generated by interpolation.

・ Determining unit 24
For any region on the image, determine whether any of the virtually generated intermediate image frames has a difference greater than the allowable value from the actual intermediate image frame based on a predetermined criterion. To do.
-Output unit 26
Encoded data including at least one of the both end image frames and the corresponding point information is output. When it is determined that there is an area having a difference greater than or equal to the allowable value, difference information regarding the area is output together.

[2] Image decoding technique The image decoding technique according to the first embodiment operates to decode the encoded data generated by the image encoding technique of [1]. Therefore, an image encoding / decoding system having a combination of the technique [1] and the following technique is a modification of the embodiment. Hereinafter, description will be made with the whole serial number attached.

  FIG. 22 shows the flow of the present image decoding technique, and at the same time, shows the configuration of an image decoding apparatus to be described later.

(13) An aspect of the image decoding method according to the embodiment executes the following steps.
p) Input encoded data in a format including at least F0, M0-n, and predetermined difference information. For example, the predetermined difference information is “iv) difference information (image data format)” in (6).
q) A path for moving the point p0 on F0 to the corresponding point pn on Fn by M0-n is divided into n points, the point p1 corresponding to p0 on F1, the point p2 corresponding to p0 on F2, Calculate a point pn-1 corresponding to p0 on Fn-1. Same as step b.

r) By executing step q for a predetermined number of points on F0, a virtual F1 ′ using a set of points p1 corresponding to the predetermined number of points and a virtual using a set of points p2 F2 ′,..., And virtual pn ′ are generated using the set of points pn. Same as step c.
s) The difference among the set S1 of the virtual F1 ′ and the real F1, the set S2 of the virtual F2 and the real F2,..., and the set Sn of the virtual Fn and the real Fn. Sk (k = 1,..., N) to which information is given is specified. This identification is performed by, for example, (iii) “iii)“ existence / non-existence bit ”indicating presence / absence of difference information and / or“ v) k value (k1, k2,.

t) A corrected virtual Fk ″ is generated by adding the difference determined by the difference information to the virtual Fk ′. In addition of the difference, reference is made to “vi) region position / shape information (A1 (k1), A2 (k2),... In a plurality of cases)” in (6).
u) As decoding results, F0, virtual F1 ′, virtual F2 ′,..., modified virtual Fk ″, virtual Fk + 1 ′,..., virtual Fn−1 'Is output in this order, that is, in the display order. As for Fn, the virtual one may be output, but when actual Fn exists, it may be output. The output destination is, for example, a display control unit, and the unit converts the display device. The image decoding apparatus according to an aspect of the embodiment includes such a display control unit and a display apparatus.

(14) The difference information describes the difference only for the region where the difference between the image frames is large, and step t specifies the position information of the region when adding the difference. This has already been described.
(15) The difference information is compressed, and step t may be added after the difference information is expanded.
(16) M0-n may be generated in a format indicating a trajectory where p0 reaches pn via pk. That is, M0-n here is not the original M0-n but a modified M0-n generated on the encoding side. According to this aspect, the image quality is improved.

(17) The image decoding apparatus according to the embodiment includes the following configuration.
・ Input unit 30
Encoded data in a format including at least F0, M0-n, and predetermined difference information is input. The input unit may be any interface, or may be a read control unit that reads out from a memory in which encoded data is stored.
Intermediate frame generation unit 32
A path for moving the point p0 on F0 to the corresponding point pn on Fn by M0-n is divided into n points, the point p1 corresponding to p0 on F1, the point p2 corresponding to p0 on F2,. By executing a process of calculating a point pn-1 corresponding to p0 on Fn-1 for a predetermined number of points on F0, a virtual set using the set of points p1 corresponding to the predetermined number of points is used. A virtual Fn ′ is generated using a set of F1 ′, point p2, and a virtual F2 ′,..., Point pn, respectively.

-Specific part 34
The difference between the set S1 of the virtual F1 ′ and the real F1, the set S2 of the virtual F2 ′ and the real F2,..., And the set Sn of the virtual Fn ′ and the real Fn. Sk (k = 1,..., N) to which information is given is specified. Examples of specific methods have already been mentioned.
Intermediate frame correction unit 36
A corrected virtual Fk ″ is generated by adding the difference determined by the difference information to the virtual Fk ′.
・ Output unit 38
As a decoding result, F0, virtual F1 ′, virtual F2 ′,..., Modified virtual Fk ″, virtual Fk + 1,..., Virtual Fn−1 ′ are output. To do. The output destination may be a display control unit that generates data or signals for the display device.
The present apparatus may further include this display control unit, and may further include the display apparatus itself.

(18) Another aspect of the image decoding method according to the embodiment executes the following processing.
Process 1: Encoded data including one of both end image frames of an image group including three or more image frames, corresponding point information between the both end image frames, and predetermined difference information is input.
Process 2: Based on the corresponding point information, an intermediate image frame sandwiched between both end image frames is virtually generated by interpolation.
Process 3: On the image of the intermediate image frame of the set described in the encoded data as a set having a large difference among the sets of the virtually generated intermediate image frames and the corresponding actual intermediate image frames To identify areas with large differences.
Process 4: A corrected virtual image frame is generated by adding the difference in the region to a virtual image frame included in a set having a large difference.
Process 5: As decoding results, one of the two-end image frames, a virtual intermediate image frame that has been modified for a set with a large difference, and a virtual intermediate image frame for the other set are used as decoded data. Output.

(19) Another aspect of the image decoding apparatus according to the embodiment includes the following configuration.
・ Input unit 30
Encoded data including one of both end image frames of an image group including three or more image frames, corresponding point information between the both end image frames, and predetermined difference information is input.
Intermediate frame generation unit 32
Based on the corresponding point information, an intermediate image frame sandwiched between both end image frames is virtually generated by interpolation.

-Area specifying part 34 '
Among the sets of virtually generated intermediate image frames and the corresponding actual intermediate image frames, the difference is present on the intermediate image frame image of the set described in the encoded data as a set having a large difference. Identify large areas. Examples of specific methods have already been described.
Intermediate frame correction unit 36
A corrected virtual image frame is generated by adding the difference in the region to a virtual image frame included in a set having a large difference.
・ Output unit 38
As a decoding result, one of the two-end image frames, a virtual intermediate image frame that has been corrected for a set with a large difference, and a virtual intermediate image frame for the other set are output as decoded data. Here again, there is a similar modification (17).

(20) Each step of the image encoding method described in (11) or the image encoding method described elsewhere may be executed by a computer program by a computer program.
(21) Each step of the image decoding method described in (18) or the image decoding method described elsewhere may be executed by a computer using a computer program.

(22) The image processing system according to the embodiment includes an image encoding unit (100 in FIG. 19) and an image decoding unit (200 in FIG. 22). This system can be used as, for example, a moving image recording / playback apparatus using a hard disk. First, the image encoding unit has the following configuration.
・ Matching processing unit 20
Matching calculation is performed between the two end image frames of the image group including three or more image frames.
Encoding side intermediate frame generation unit 22
Based on the corresponding point information between the two-end image frames obtained as a result of the matching calculation, an intermediate image frame sandwiched between the two-end image frames is virtually generated by interpolation.
・ Determining unit 24
For any region on the image, determine whether any of the virtually generated intermediate image frames has a difference greater than the allowable value from the actual intermediate image frame based on a predetermined criterion. To do.

.Write control unit (26)
Encoded data including at least one of the both-end image frames and the corresponding point information is written into a memory (not shown). When it is determined that there is an area having a difference equal to or greater than the allowable value, difference information regarding the area is also written.

On the other hand, the image decoding unit has the following configuration.
Read control unit (30)
Read the encoded data from the memory.
Decoding side intermediate frame generation unit 32
Based on the image frame data and the corresponding point information included in the encoded data, an intermediate image frame sandwiched between both end image frames is virtually generated by interpolation. Note that the same configuration as that of the encoding-side intermediate frame generation unit may be shared. In that case, a selector may be provided that selects either the output of the encoding-side matching processing unit or the output of the decoding-side readout control unit and inputs the selected output to the intermediate frame generation unit. This selector selects the output of the matching processing unit at the time of encoding and the output of the reading control unit at the time of decoding.

-Area specifying unit 34
An area in which the difference information is included in the encoded data is specified. Examples of the method have already been described.
Intermediate frame correction unit 36
A corrected virtual image frame is generated by adding the difference in the region to the virtual image frame including the specified region.

・ Output unit 38
As a decoding result, one or both of the two end image frames, a virtual intermediate image frame modified for the virtual image frame including the region, and a virtual image frame not including the region The virtual intermediate image frame itself is output as decoded data.

(Second Embodiment)
Hereinafter, an encoding technique and a decoding technique according to the second embodiment will be described in order. FIG. 23 shows the configuration of the image encoding apparatus according to the second embodiment, and at the same time, shows the flow of the image encoding technique.

[1] Configuration of Encoding Device CPF: Image Matching Processor that Uses Critical Point Filter as a Premise Technology, that is, Singularity Filter Matching between key frames is calculated for each pixel, and corresponding point information is output. This information is output as a file. This file describes which pixel of the source side key frame corresponds to each pixel of the destination side key frame. Therefore, a morphing image between two key frames can be obtained by interpolating pixel positions and pixel values corresponding to each other between these key frames based on this file. Note that if this file is applied only to the source side key frame and the interpolation calculation is performed, a morphing image is obtained in which each pixel of the source side key frame is gradually moved to the position of the corresponding pixel described in this file. It is done. In this case, only the position is interpolated between the corresponding pixels.

  Although an image matching processor can be widely used instead of CPF, pixel matching with high accuracy is ideal for the purpose of the present embodiment, and the prerequisite technology satisfies the condition.

  DE: Differential Encoder A differential (error) encoder. The difference between two image frames is variable-length coded based on Huffman coding or other statistical methods.

  NR: maskable Noise Reducer Noise reducer. In many cases, human vision cannot recognize minute changes. For example, in a portion where the luminance changes drastically, that is, in a region where a component having a high luminance spatial frequency component is strong, the luminance variation error is not visually grasped. Noise is superimposed on the moving image information in various forms, and such data is merely recognized as noise visually and has no meaning as an image. Ignoring such visually insignificant random information, ie “visual mask information”, is important to achieve higher compression ratios.

  The quantization in the current block matching uses visual mask information about luminance values, but there are some visual mask information besides the luminance values. NR utilizes a visual mask for spatial location information as well as temporal location information. The spatial mask of the spatial position information uses the fact that the phase component of the spatial frequency is difficult to visually recognize in the case of an image with a complicated luminance change. The visual mask of the time position information utilizes the fact that even if the change in the data in the time direction is shifted in the portion where the change in the time direction is severe, the difference is not easily recognized visually. These are detected by comparing with a predetermined threshold value.

  At least in the current MPEG scheme of block matching and differential encoding, it is difficult to actively use these masks. On the other hand, the decoding process in the base technology generates a change on the moving image by trilinear or other interpolation in order to avoid discontinuity that causes visual unnaturalness. Not only the direction but also the function of making it visually inconspicuous by being scattered in the space direction and the time direction. NR is useful in combination with the underlying technology.

  DD: Differential Decoder A differential (error) decoder. The difference encoded by DE is decoded and added to the image frame in which the difference occurs, thereby improving the accuracy of the image frame.

  DC: Differential Comparator Virtual F1 ′ and real F1 set S1, virtual F2 and real F2 set S2,..., Virtual Fn and real Fn set Sn included in the set The presence / absence of a set Sk (k = 1,..., N) having a large difference between image frames is determined based on a predetermined determination criterion.

  In addition to these, there is a function of causing corresponding point information to act on a single key frame and virtually generating another key frame only from pixel movement of the key frame. Hereinafter, a functional block that realizes this function is referred to as a pixel shifter.

  Note that the CPF and DC in the second embodiment can correspond to the matching processing unit 20 and the determination unit 24 in the first embodiment, respectively.

[2] Encoding process In FIG. 23, “F0” and the like indicate each frame of the moving image to be processed, and “M0-n” indicates corresponding point information between F0 and Fn generated by the CPF. Encoding proceeds in the following procedure. Hereinafter, a case where n = 4 will be described. In FIG. 23, n ₂ = 8.

a) Matching is calculated by CPF between the first and second key frames (F0, F4) sandwiching one or more image frames (F1 to F3), and corresponding point information between the first and second key frames ( Generating M0-4).
b-1) A step of generating a correction M0-4 ′ based on the corresponding point information (M0-4) between the first and second key frames. For generation of the correction M0-4 ′, the technique described in the first embodiment may be used.
b-2) Based on the corrected corresponding point information (M0-4 ′) between the first and second key frames, the pixels included in the first key frame (F0) are moved by the pixel shifter to create a virtual Generating a second key frame (F4 ′);
c) The step of compressing and encoding the difference between the actual second key frame (F4) and the virtual second key frame (F4 ′) with DE with NR function (denoted as DE + NR).
d) First key frame (F0), modified corresponding point information (M0-4 ′) between the first and second key frames, and compression between the actual second key frame and the virtual second key frame Outputting the encoded difference (Δ4) as encoded data between these key frames (F0, F4); The output destination may be a recording medium or a transmission medium. In practice, it is integrated with the information output in j) described later, and is output to a recording medium or the like as moving image encoded data.

Step b-1) will be described in detail. As described in FIG. 19, a set Sk (k = 1,..., N) having a large difference between image frames included in the set from the difference Si between the actual frame Fi and the corresponding virtual frame Fi ′. Is determined based on a predetermined criterion.
The determination result is output to DE + NR, the difference of the pair Sk is compressed, and is output as Δk as difference information.

  In addition, a value of k indicating a set in which difference information exists is output to the CPF. The CPF calculates corresponding point information between adjacent frames between the frames F0 to Fk. That is, corresponding point information (M0-1, M1-2, M2-3,...) Between F0 and F1, F1 and F2, F3 and F4,. The combiner CONCAT combines these corresponding point information and outputs corresponding point information M0-k. Corresponding point information M0-k is combined with corresponding point information M0-k generated in step a) to generate corrected corresponding point information M0-n '.

Subsequently, the following processing is performed for the second key frame (F4) and thereafter.
e) Decoding the difference (Δ4) compression-coded between the actual second key frame (F4) and the virtual second key frame (F4 ′) with DD.
f) Generating an improved virtual second key frame (F4 ″) by DD from the decrypted difference and the virtual second key frame (F4 ′).
g) Matching is calculated by CPF between the second and third key frames (F4, F8) sandwiching one or more image frames (F5 to F7), and corresponding point information between the second and third key frames ( Generating M4-8).
h-1) A step of generating (M4-8 ′) based on the corresponding point information (M4-8) between the second and third key frames.
h-2) Based on the modified corresponding point information (M4-8 ′) between the second and third key frames, it is included in the virtual second key frame (F4 ″) improved by the pixel shifter. Generating a virtual third key frame (F8 ′) by moving the pixels;
i) A step of compressing and encoding the difference between the actual third key frame (F8) and the virtual third key frame (F8 ′) with DE + NR.
j) The corrected corresponding point information (M4-8 ′) between the second and third key frames, and the difference (Δ8) compression-coded between the actual third key frame and the virtual third key frame. A step of outputting as encoded data between these key frames (F4, F8). The output destination is generally the same as the output destination of d).

  Thereafter, the subsequent steps e) to j) are sequentially repeated for subsequent key frames, and the repetition process is terminated when a predetermined group end key frame is reached. The group end key frame corresponds to a 1 GOP end frame in MPEG. Therefore, the next frame after this frame is newly regarded as the first key frame as the first frame of the new group, and a) the following processing is repeated. With the above processing, for a group corresponding to a GOP in MPEG (hereinafter simply referred to as a group), only one image corresponding to a key frame (I picture in MPEG) needs to be encoded and transmitted.

[3] Configuration of Decoding Device FIG. 24 is a diagram illustrating a flow of the image decoding technique according to the second embodiment and a configuration of the image decoding device.
It is a simpler configuration than the encoding side.
DD: Same as DD of encoding device.
INT: INTerpolator interpolation processor.

In addition to these, there is a pixel shifter similar to the encoding side. An intermediate frame is generated by interpolation processing from the two image frames and corresponding point information.
[4] Decoding process Decoding proceeds in the following procedure. Here, it is assumed that n = 4 and n ₂ = 8.

k) Modified corresponding point information (M0-4 ′) between the first and second key frames (F0, F4) sandwiching one or more image frames (F1 to F3), and the first key frame (F0) ) Step to get. Acquisition may be from either a transmission medium or a recording medium.
l) Based on the corrected corresponding point information (M0-4 ′) between the first and second key frames, by moving the pixels included in the first key frame (F0) by the image shifter, a virtual Generating a second key frame (F4 ′);
m) On the encoding side in advance l) By the same processing, a virtual second key frame (F4 ′) is generated, and the encoding side compresses the difference between this and the actual second key frame (F4). Since the digitized data (Δ4) is generated, this is acquired.
o) The obtained compressed encoded data (Δ4) of the difference is decoded by DD and added to the virtual second key frame (F4 ′) to obtain the improved virtual second key frame (F4 ″). Step to generate.
p) Based on the corrected corresponding point information (M0-4 ′) between the first and second key frames, the first key frame (F0) and the virtual second key frame (F4) improved by INT ") To generate intermediate frames (F1 'to F3') that should exist between these key frames (F0, F4") by performing an interpolation calculation between them.
q) The first key frame (F0), the generated intermediate frames (F1 ′ to F3 ′), and the improved virtual second key frame (F4 ″) are decoded data between these key frames to a display device or the like. Step to output.

Subsequently, the following processing is performed for the second key frame (F4) and thereafter.
r) obtaining corrected corresponding point information (M4-8 ′) between the second and third key frames (F4, F8) sandwiching one or more image frames (F5 to F7).
s) Based on the modified corresponding point information (M4-8 ′) between the second and third key frames, the pixels included in the improved virtual second key frame (F4 ″) are converted by the pixel shifter. Generating a virtual third key frame (F8 ′) by moving.
t) In the same manner, s) on the encoding side in advance generates a virtual third key frame (F8 ′) on the encoding side, and this and the actual third key frame (F8) on the encoding side. The step of generating differentially encoded data (Δ8) and acquiring it.
u) A step of generating an improved virtual third key frame (F8 ″) by DD from the obtained compressed encoded data (Δ8) of the difference and the virtual third key frame (F8 ′).
v) On the basis of the modified corresponding point information (M4-8 ′) between the second and third key frames, an improved virtual second key frame (F4 ″) and an improved virtual Generating an intermediate frame (F5′-F7 ′) to be present between these key frames by performing an interpolation calculation between the third key frames (F8 ″);
w) The improved virtual second key frame (F4 ″), the generated intermediate frames (F5 ′ to F7 ′), the improved virtual third key frame (F8 ″) are transferred to these key frames (F4 ″). , F8 ″) is output as decoded data to a display device or the like.

Thereafter, steps r) to w) are sequentially repeated for the subsequent key frames, and the repetition process is terminated when the group end key frame is reached. The next frame of this frame is newly regarded as the first key frame as the first frame of the new group, and k) the following processing is repeated.
[5] Advantages of this embodiment

  According to the encoding / decoding technique according to the second embodiment, the following merits can be enjoyed in addition to the encoding / decoding technique according to the first embodiment.

  When using the CPF of the base technology for image matching, since the matching accuracy is high, the compression rate realized in the present embodiment is high. This is because the difference to be compressed by DE + NR is small from the beginning and the statistical bias is large.

  Similarly, when CPF is used, since this encoding method does not use block matching, even if the compression rate is increased, there is no block noise which is a problem in MPEG. Of course, a processing method without block noise may be employed for image matching other than CPF.

  Originally, MPEG only considers minimization of differences, but CPF detects a portion that should be originally handled, so that ultimately, a higher compression ratio than MPEG can be realized.

  The encoding device can be composed of an image matching processor, a differential encoder with a noise reduction function, a differential decoder, and a pixel shifter, and is simple. Further, the noise reduction function is an optional function, and this may not be necessary. Similarly, the decoding apparatus can be composed of an interpolation processor, a differential decoder, and a pixel shifter, and is simple. In particular, the decoding device does not need to perform image matching, and the processing amount is light.

  Every time a virtual key frame is generated, the difference between the virtual key frame and the actual key frame is taken into the encoded data as Δ4, Δ8, etc., so that only one complete key frame is encoded per group. Regardless, there is no error accumulation when playing a long movie.

[6] Deformation technology

  When generating the corresponding point information file by performing matching calculation between the first and second key frames (F0, F4), intermediate frames (F1 to F3) existing between the key frames may be taken into consideration. In that case, the CPF calculates matching for each set of F0 and F1, F1 and F2, F2 and F3, and F3 and F4, and generates four files (provisionally called partial files M0 to M3). Subsequently, these four files may be integrated and output as one corresponding point information file.

  For integration, first, it is specified where each pixel of F0 moves on F1 by M0. Subsequently, it is specified where the pixel specified on F1 moves on F2 by M1. If this is performed up to F4, the correspondence between F0 and F4 becomes more accurate by the four partial files. This is because F0 and F4 have some distance, and the matching accuracy between adjacent image frames is generally higher than that between them.

  Although this method ultimately improves the matching accuracy between F0 and F4, the corresponding point information file may be expressed as a function of time. In this case, the partial files are not integrated, and the four files may be regarded as corresponding point information files and provided to the decoding side. The decoding side can decode a more accurate moving image by iterative processing of generating F1 from F0, F4, and M0 and generating F2 from F0, F4, M0, and M1.

(Third embodiment)
Another embodiment of the present invention relates to the encoding apparatus of FIG. Here, image matching energy is introduced as a measure indicating the accuracy of image matching, and this is used for noise reduction or the like in DE + NR. Hereinafter, description will be made with reference to FIG. 23 as appropriate, but the configurations and functions not particularly mentioned are the same as those in the second embodiment.

  The matching energy here is determined by the difference between the distance between corresponding points and the pixel value, and is represented by, for example, Expression 49 in the base technology. In the present embodiment, this matching energy obtained at the time of image matching in the CPF is used as a by-product. In the base technology image matching, for each pixel between key frames, a pixel having the minimum mapping energy is detected as a corresponding point. Focusing on these features of the base technology, good matching is achieved for pixels with low matching energy, while pixels with large changes in position and pixel values between key frames are naturally found at locations with high matching energy. However, in some cases, it can be evaluated that there may be a matching error. Although described in detail below, in the present embodiment, the compression rate of the difference is increased for a portion with high matching accuracy. In another example, difference information relating to a pixel for which a matching error is estimated may be highly compressed.

[1] Encoding process In the encoding apparatus according to the present embodiment, when the CPF calculates the matching between the first and second key frames, the matching energy of each pixel corresponding to both frames is acquired together. Then, an energy map describing the matching energy of each pixel is generated on the first key frame (F0). Similarly, energy maps are generated between other adjacent key frames. That is, the energy map is data in which the matching energy of each corresponding point between key frames is basically described for each pixel of the previous key frame. Note that the energy map may be represented on the subsequent key frame among the previous and next key frames. The energy map is sent from the CPF to the DE + NR by a route not shown. In DE + NR, this energy map is used to evaluate the quality of matching between key frames, and based on this, the difference between a virtual key frame and an actual key frame is adaptively compressed and encoded. In addition to the energy map, a corresponding point information file is also sent to DE + NR via a route (not shown).

  FIG. 25 is a diagram showing a configuration of DE + NR in FIG. 23 according to the present embodiment. 25 includes a difference calculator 10, a difference compression unit 12, an energy acquisition unit 14, and a determination unit 16. Of these, the former two correspond exclusively to DE and the latter two exclusively correspond to NR. The operation of DE + NR when encoding the first key frame (F0), the second key frame (F4), and the intermediate image frames (F1 to F3) will be described below. The DE + NR operation is the same in the frame coding.

  The difference calculator 10 obtains the actual second key frame (F4) and the virtual second key frame (F4 '), and obtains the difference between the pixel values of the pixels that correspond in position. Thereby, a kind of image in which each pixel has a difference in pixel value between both key frames is formed, and this is called a difference image. The difference image is sent to the energy acquisition unit 14. Further, the energy acquisition unit 14 receives an energy map and corresponding point information (M0-4) between the actual first key frame (F0) and the actual second key frame (F4) from the CPF in FIG. Is done. The energy acquisition part 14 acquires the matching energy of a difference image using these.

  First, the acquisition unit 14 acquires corresponding point information (M0-4) between the first and second key frames from the CPF. Using this, by tracing the difference image from the virtual second key frame (F4 ′) and the first key frame (F0), which pixel of the difference image is which pixel of the first key frame (F0) Corresponding to the shifted one or the correspondence is acquired. Then, the energy of each pixel on the energy map represented on the first key frame is referred to, and the matching energy of the pixel on the first key frame (F0) corresponding to each pixel of the difference image is determined as the difference image. Obtained as matching energy for each pixel. The matching energy of the difference image is thus obtained.

  The energy acquisition unit 14 sends the matching energy of the difference image to the determination unit 16. The determination unit 16 uses the matching energy of each pixel of the difference image to determine a region to be highly compressed in the difference image, and notifies the compression unit 12 of information on which region should be highly compressed. The determination is performed as follows, for example. The determination unit 16 divides the difference image into blocks of 16 × 16 pixels, and compares the matching energy with a predetermined threshold value for all the pixels included in each block. As a result of the comparison, if the matching energy of all the pixels in the block is equal to or less than the threshold value, the area is determined as a high compression target block.

  The compression unit 12 compresses the difference image in the JPEG format. At this time, the compression rate is adaptively changed between the normal region and the high compression compatible region using the information of the high compression compatible region notified from the determination unit 16. Specifically, a process for increasing the quantization width of the DCT coefficient compared to a normal block can be used for the high compression target block. In another example, in the difference image, the JPEG compression may be performed after the pixel value of the high compression target block is set to 0. In any case, the reason why the region where the matching energy is low is highly compressed is as follows.

  That is, as described above, a pixel having a low matching energy can be regarded as having a good matching result between key frames. Therefore, the difference between the actual second key frame (F4) and the virtual second key frame (F4 ′) is unlikely to occur in the portion of the difference image where the matching energy is low. If so, it may be considered noise. Therefore, an area with low matching energy in the difference image can be compressed much more than other areas without worrying about missing information due to high compression. On the other hand, in a region where the matching energy is large, an error may occur in matching, and the difference between the virtual second key frame (F4 ′) and the actual second key frame (F4) is important information for decoding. Therefore, the compression rate is kept low and priority is given to the accuracy at the time of decoding.

[2] Merits of Third Embodiment Through the above processing, the compression unit 18 performs the compression-coded difference between the actual second key frame (F4) and the virtual second key frame (F4 ′) ( Δ4) is output. According to the encoding apparatus according to the present embodiment, the difference information between the actual key frame and the virtual key frame is adaptively adapted according to the importance of accurately decoding the encoded image more faithfully to the original image. Therefore, high encoding efficiency can be realized while maintaining decoding accuracy. Needless to say, this embodiment can also benefit from the advantages of the first embodiment.

[3] Modified Technology of Third Embodiment As a modified example of the present embodiment, a pixel having a large matching energy, particularly a pixel that exhibits a corresponding tendency that is significantly different from the corresponding tendency of neighboring pixels may cause a matching error. Since it is empirically recognized that there are many, it is possible to evaluate a pixel whose matching energy is significantly different from that of surrounding pixels as a matching error, and to introduce this into noise reduction. In this case, DE + NR compares the matching energy of each pixel of the second key frame (F4) with, for example, the average of the matching energy of other pixels in a 9 × 9 pixel block centered on itself. When the difference between the two exceeds a predetermined threshold as a result of the comparison, it may be determined that such a pixel has caused a matching error.

  Corresponding information causing an error can be considered as meaningless data for the decoding side, and in the difference information between the actual second key frame (F4) and the virtual second key frame (F4 ′), Data relating to a pixel causing a matching error can be said to be noise. Therefore, it is unnecessary to consider the loss of information due to high compression, and DE + NR is a pixel that corresponds to a matching error between actual key frames among the difference images between actual key frames and virtual key frames. Compress at a higher rate than The determination of the matching error is performed, for example, by comparing the tendency of the motion vector of the surrounding pixels and the tendency of the motion vector of the pixel of interest, and determining whether the motion vector of the pixel of interest is significantly different from the surrounding tendency. May be.

  Also in the third embodiment, as in the second embodiment, the intermediate frames (F1 to F3) between the first and second key frames (F0, F4) are taken into consideration, and each of these image frames adjacent to each other is considered. There is a modification technique in which matching is calculated for a pair, corresponding point information files (M0 to M3) are generated, and these are integrated to obtain one corresponding point information file between the first and second key frames (F0, F1). Conceivable. Like the modification technique of the first embodiment, matching accuracy can be improved and accurate video decoding can be realized.

  Furthermore, with this modification technique, the matching energy between the image frames can be calculated and applied to scene change detection and the like. The configuration related to scene change detection is as follows. First, CPF performs matching calculation for each set of F0 and F1, F1 and F2, F2 and F3, F3 and F4, etc., and obtains energy maps, E0, E1, E2, E3,. To do. Here, the average of the matching energy relating to the pixels of an entire image frame is taken and compared with a predetermined threshold for scene change detection, and the image immediately after that is taken as a new group. For example, it is assumed that, as a result of averaging the matching energy of each pixel of F5 related to the matching of F5 and F6 based on the energy map E5 between F5 and F6, the value exceeds the key frame addition threshold value. In this case, the immediately following key frame, that is, F6 and below may be a new group, and F6 may be the first key frame of the next group. This is because when the matching energy is large, it can be considered that there is a large change between images. Thereby, an automatic scene change can be detected, and a group can be selected corresponding to the scene change.

  Based on each energy map, the average matching energy of the pixels in each image frame is calculated and cumulatively added. When that value exceeds a predetermined threshold, the image frame is newly added. It may be registered as a key frame. This is because if the key frame can be added when the cumulative amount of change between image frames exceeds a certain value, the image quality at the time of decoding can be further improved.

  The present invention can be used in the field of image compression processing technology.

Claims (22)

A sequence of image frames is expressed as F0, F1,..., Fn-1, Fn (n is an integer of 2 or more) in descending or ascending order, and image frames Fi, Fj (i, j = 0, 1,... .., N) When the corresponding point information indicating the positional relationship of the points corresponding to each other is expressed as Mi-j,
a) calculating a match between F0 and Fn to generate M0-n;
b) A path for moving the point p0 on F0 to the corresponding point pn on Fn by M0-n is divided into n points, points p1 corresponding to p0 on F1, points p2 corresponding to p0 on F2, Calculating a point pn corresponding to p0 on Fn;
c) By executing step b for a predetermined number of points on F0, a virtual F1 using a set of points p1 corresponding to the predetermined number of points and a virtual set using points p2 is used. F2,..., Generating virtual Fn using a set of points pn,
d) A set S1 of virtual F1 and real F1, a set S2 of virtual F2 and real F2,..., and a set Sn of virtual Fn and real Fn. Determining the presence or absence of a set Sk (k = 1,..., N) having a large difference between included image frames based on a predetermined determination criterion;
e) When there is a set Sk having a large difference, at least a step of calculating matching between Fh (h = 0, 1,..., k−1) and Fk to generate Mh−k;
f) using the information of Mh-k to correct M0-n;
g) outputting encoded data in a format including at least F0 and the modified M0-n;
An image encoding method comprising:   2. The image encoding method according to claim 1, wherein step d determines the magnitude of the difference between image frames in units of a predetermined area. The method of claim 1, wherein step e comprises
e1) When there is a set Sk having a large difference, M0-1 between F0 and F1, M1-2 between F1 and F2, ..., M (k-1) -k between Fk-1 and Fk Each step to seek,
e2) integrating M0-1, M1-2, ..., M (k-1) -k to generate M0-k;
An image encoding method comprising:   The method according to claim 3, wherein step e2 obtains p1 corresponding to p0 by M0-1, obtains p2 corresponding to p1 by M1-2, ..., by M (k-1) -k. pk corresponding to pk-1 is obtained, and points corresponding to p0 are specified in order of p1, p2,..., pk, and finally pk corresponding to p0 is specified to generate M0-k. An image encoding method characterized by the above.   5. The method of claim 4, wherein step f uses the information of M0-k to generate a modified M0-n in a form that indicates a trajectory where p0 passes through pk to pn. An image encoding method to be performed.   The method according to claim 1, wherein if it is found in step d that a set Sk having a large difference exists, step g includes a difference information in the set Sk in addition to F0 and M0-n modified. An image encoding method characterized in that the encoded data is output.   7. The image encoding method according to claim 6, wherein the difference information in the set Sk is included in the encoded data only for a region having a large difference in the image frame.   8. The method according to claim 7, wherein at least a value of k and position information of the region are included in the encoded data together with the difference.   9. The image encoding method according to claim 8, wherein the difference is included in the encoded data after being subjected to compression processing. A sequence of image frames is expressed as F0, F1,..., Fn-1, Fn (n is an integer of 2 or more) in descending or ascending order, and image frames Fi, Fj (i, j = 0, 1,... .., N) When the corresponding point information indicating the positional relationship of the points corresponding to each other is expressed as Mi-j,
A matching processing unit that calculates matching between F0 and Fn and generates M0-n;
A path for moving the point p0 on F0 to the corresponding point pn on Fn by M0-n is divided into n points, the point p1 corresponding to p0 on F1, the point p2 corresponding to p0 on F2,. By executing a process for calculating a point pn corresponding to p0 on Fn for a predetermined number of points on F0, a virtual F1, a point p2 using a set of points p1 corresponding to the predetermined number of points An intermediate frame generation unit that generates a virtual Fn using a set of virtual F2,...
A set S1 of virtual F1 and real F1, S2 of virtual F2 and real F2,..., And a set Sn of virtual Fn and real Fn are included in the set. A determination unit that determines the presence or absence of a set Sk (k = 1,..., N) having a large difference between image frames based on a predetermined determination criterion;
The matching processing unit calculates Mh−k by calculating a matching between at least Fh (h = 0, 1,..., K−1) and Fk when a set Sk having a large difference exists. And M0-n is corrected using the information of Mh-k, and the apparatus further includes an output unit that outputs encoded data in a format including at least F0 and the corrected M0-n. An image encoding device. Performing a matching calculation between both end image frames of an image group including three or more image frames;
A step of virtually generating an intermediate image frame sandwiched between both end image frames based on the corresponding point information between the both end image frames obtained as a result of the matching calculation by interpolation,
For any region on the image, determine whether any of the virtually generated intermediate image frames has a difference greater than the allowable value from the actual intermediate image frame based on a predetermined criterion. A determination step to:
Encoded data including at least one of the two-end image frames and corresponding point information is generated, and if it is determined in the determination step that there is a region having a difference greater than or equal to an allowable value, Generating differential information in the encoded data; and
An image encoding method comprising: A matching processing unit for performing a matching calculation between both end image frames of an image group including three or more image frames;
Based on the corresponding point information between the two end image frames obtained as a result of the matching calculation, an intermediate frame generating unit that virtually generates an intermediate image frame sandwiched between the two end image frames by interpolation, and
For any region on the image, determine whether any of the virtually generated intermediate image frames have a difference greater than the allowable value from the actual intermediate image frame based on a predetermined criterion. A determination unit to perform,
Output encoded data including at least one of the both-end image frames and the corresponding point information, and if the determination unit determines that there is a region having a difference greater than or equal to an allowable value, An output unit for outputting difference information;
An image encoding device comprising: A sequence of image frames is expressed as F0, F1,..., Fn-1, Fn (n is an integer of 2 or more) in descending or ascending order, and image frames Fi, Fj (i, j = 0, 1,... .., N) When the corresponding point information indicating the positional relationship of the points corresponding to each other is expressed as Mi-j,
p) inputting encoded data in a format including at least F0, M0-n and predetermined difference information;
q) A path for moving the point p0 on F0 to the corresponding point pn on Fn by M0-n is divided into n points, the point p1 corresponding to p0 on F1, the point p2 corresponding to p0 on F2, Calculating a point pn-1 corresponding to p0 on Fn-1;
r) By executing step q for a predetermined number of points on F0, a virtual F1 using a set of points p1 corresponding to the predetermined number of points and a virtual using a set of points p2 F2,..., Generating virtual Fn using a set of points pn,
s) The difference information among a set S1 of a virtual F1 and a real F1, a set S2 of a virtual F2 and a real F2, ..., a set Sn of a virtual Fn and a real Fn Specifying Sk (k = 1,..., N) to which
t) generating a modified virtual Fk by adding a difference determined by the difference information to the virtual Fk;
u) outputting F0, virtual F1, virtual F2,..., modified virtual Fk, virtual Fk + 1,..., virtual Fn-1 as decoding results; ,
An image decoding method comprising:   14. The method according to claim 13, wherein the difference information describes a difference only for a region having a large difference between image frames, and step t specifies position information of the region when adding the difference. An image decoding method.   14. The image decoding method according to claim 13, wherein the difference information is compressed, and step t adds the difference information after expanding the difference information.   14. The image decoding method according to claim 13, wherein M0-n is generated in a format indicating a trajectory where p0 reaches pn via pk. A sequence of image frames is expressed as F0, F1,..., Fn-1, Fn (n is an integer of 2 or more) in descending or ascending order, and image frames Fi, Fj (i, j = 0, 1,... .., N) When the corresponding point information indicating the positional relationship of the points corresponding to each other is expressed as Mi-j,
An input unit for inputting encoded data in a format including at least F0, M0-n, and predetermined difference information;
A path for moving the point p0 on F0 to the corresponding point pn on Fn by M0-n is divided into n points, the point p1 corresponding to p0 on F1, the point p2 corresponding to p0 on F2,. By executing a process of calculating a point pn-1 corresponding to p0 on Fn-1 for a predetermined number of points on F0, a virtual set using the set of points p1 corresponding to the predetermined number of points is used. F1, an intermediate frame generation unit that generates a virtual Fn using a set of points pn, a virtual F2,.
The difference information is given by each of a pair S1 of a virtual F1 and a real Fn, a pair S2 of a virtual F2 and a real F2,... A specifying unit that specifies Sk (k = 1,..., N)
An intermediate frame correction unit that generates a corrected virtual Fk by adding a difference determined by the difference information to the virtual Fk;
An output unit that outputs F0, virtual F1, virtual F2,..., Modified virtual Fk, virtual Fk + 1,.
An image decoding apparatus comprising: Inputting encoded data including one of both end image frames of an image group including three or more image frames, corresponding point information between the both end image frames, and predetermined difference information;
Based on the corresponding point information, virtually generating an intermediate image frame sandwiched between both end image frames by interpolation; and
Among the sets of virtually generated intermediate image frames and the corresponding actual intermediate image frames, the difference is present on the intermediate image frame image of the set described in the encoded data as a set having a large difference. Identifying a large area;
Generating a corrected virtual image frame by adding a difference in the region to a virtual image frame included in a set having a large difference; and
A step of outputting, as decoded data, one of both-end image frames, a virtual intermediate image frame corrected for a set having a large difference, and a virtual intermediate image frame for the other set as decoded data. When,
An image decoding method comprising: An input unit that inputs encoded data including one of both end image frames of an image group including three or more image frames, corresponding point information between the both end image frames, and predetermined difference information;
Based on the corresponding point information, an intermediate frame generation unit that virtually generates an intermediate image frame sandwiched between both end image frames by interpolation, and
Of the pair of virtually generated intermediate image frames and the corresponding actual intermediate image frame, the difference is present on the intermediate image frame image of the set described in the encoded data as a set having a large difference. An area specifying unit for specifying a large area;
An intermediate frame correction unit that generates a corrected virtual image frame by adding a difference in the region to a virtual image frame included in a set having a large difference; and
As a decoding result, an output that outputs one of the two-end image frames, a virtual intermediate image frame that has been modified for a set with a large difference, and a virtual intermediate image frame for the other set as decoded data And
An image decoding apparatus comprising:   A computer program causing a computer to execute each step of the image encoding method according to claim 11.   A computer program causing a computer to execute each step of the image decoding method according to claim 18. In an image processing system having an image encoding unit and an image decoding unit,
The image encoding unit is
A matching processing unit for performing a matching calculation between both end image frames of an image group including three or more image frames;
Based on the corresponding point information between the two-end image frames obtained as a result of the matching calculation, an encoding-side intermediate frame generation unit that virtually generates an intermediate image frame sandwiched between the two-end image frames by interpolation,
For any region on the image, determine whether any of the virtually generated intermediate image frames have a difference greater than the allowable value from the actual intermediate image frame based on a predetermined criterion. A determination unit to perform,
Write control in which encoded data including at least one of both-end image frames and corresponding point information is written to the memory, and if it is determined that there is an area having a difference greater than or equal to an allowable value, difference information regarding the area is written With
The image decoding unit is
A read control unit for reading encoded data from the memory;
A decoding-side intermediate frame generation unit that virtually generates an intermediate image frame sandwiched between both-end image frames by interpolation based on image frame data and corresponding point information included in the encoded data;
An area specifying unit for specifying an area in which difference information is included in encoded data;
An intermediate frame correction unit that generates a corrected virtual image frame by adding a difference in the region to a virtual image frame including the identified region;
As a decoding result, one or both of the two end image frames, a virtual intermediate image frame modified for a virtual image frame including the region, and a virtual image frame not including the region An output unit that outputs the virtual intermediate image frame itself as decoded data;
An image processing system comprising:

Images