JP4060559B2 - Image processing apparatus and image processing method - Google Patents

Description

【００４８】
図４は上記した例で示した画素変換のカラーモデルを示し、図５はそのモデルの断面を表している。図５の中央部の小さい円の領域は、白、グレー、黒のいずれかになる領域である。
【００４９】
Ｈの値により赤(red)、黄色(yellow)、緑(green)、シアン(cyan)、青(blue)、マゼンダ(magenta)の何れかに分類される。
【００５０】
以上の処理の結果、各画素は赤(red)、黄色(yellow)、緑(green)、シアン(cyan)、青(blue)、マゼンダ(magenta)、白(white)、グレー(gray)、黒(black)の何れかになる。
【００５１】
これらの変換処理を入力された画像データの一行分行ない、その結果をメモリ内に保存する。
【００５２】
次にＨの値から更に大まかな色分布の決定を行なう（ステップＳＴ３０２）。ここでは、現在の行の色分布データと一つ前の行の色分布データからｘ方向の解像度を落としながら色分布を決定していく。
【００５３】
_ｘ方向の解像度を１／Ｎに落とし、色分布を決定する画素位置を（ｉ，ｊ）とすると、ｉ−１行で前後３Ｎ画素、ｉ行で前後３Ｎ画素分の色分布を調べ、赤(red)、黄色(yellow)、緑(green)、シアン(cyan)、青(blue)、マゼンダ(magenta)、白(white)、グレー(gray)、黒(black)ごとにカウントする。その結果、最もカウント数が多い色をその画素の色分布と決定する。
【００５４】
図６はＮ＝３とした例を示している。図中のｒは赤(red)を、ｍはマゼンダ(magenta)を、ｇはグレー(gray)を、ｂは黒(black)を表している。該当する画素に於いて各々の色をカウントすると、ｒ＝８、ｇ＝５、ｍ＝３、ｂ＝２となる。これより色分布は赤と決定できる。以上の処理により、解像度を落とした大まかな色の分布状態がわかる色分布を決定できる。
【００５５】
そして、決定した色分布から色分布画像データを生成する（ステップＳＴ３０３）。これは、ランレングスデータのように「色の識別子＋連続する画素数」という表現でデータを記述することにより、更に圧縮することが可能となる。その結果、色分布画像データ１５２が出力される。
【００５６】
以上により色分類部１０３の出力として、色分布画像データが生成される。
【００５７】
次に画素属性識別処理について説明する。
図７は本発明の第１実施形態に於ける画像処理装置に設けられた画素属性識別部１０４の画素属性識別処理手順を示すフローチャートであり、図２のステップＳＴ２０６で行われる処理の詳細を示すフローチャートである。
【００５８】
画素属性識別処理とは、領域識別と色分類の結果を比較することによって、領域識別結果の確認と修正を行なう処理である。以下、図７に示すフローチャートに従って画素属性識別処理動作を説明する。
【００５９】
領域識別部１０２では、文書要素を抽出して、識別を行ない、その結果を出力する。また、色分類部１０３では、数種類の色の分類を行ない、その結果を色分布画像データとして出力する。画素属性識別部１０４では、領域識別の結果とそれに対応する位置の色分布画像データの情報から領域識別結果の再判定を行なう。以下の例では、カラー写真と識別された領域に対しての再判定方法を述べる。尚、ここでの写真はカラー写真のことを指す。
【００６０】
まず領域識別データ１５３が写真属性かどうかを調べる（ステップＳＴ４０１）。写真属性を持っていない場合、例えばテキスト属性である場合、そのまま修正領域データ１５４として出力する。
【００６１】
ステップＳＴ４０１に於いて領域識別データ１５３が写真属性であった場合、色分布画像データ１５２から該当する領域の色分布情報のうち、色数を調査する（ステップＳＴ４０２）。写真であるならば、さまざまな色で表現されることが多い。そのような領域では、色分類の結果として、いくつもの色が検出される。逆に、色分類の結果、単色に近い場合、複雑な線図形とみなすことができる。多くの色が存在している場合、写真と判定し、そのまま修正領域データ１５４として出力する（ステップＳＴ４０３）。また、単色に近い場合、次に色分布画像データ１５２から該当する領域の色分布の偏りを調査する（ステップＳＴ４０４）。写真であるならば、色が領域中に散らばっていることが多い。逆に、線図形であるならば、オブジェクトに単一色を割り当てることが多いため、色の分布に偏りが生じることが予想される。色の偏りを調べる方法としては、色分布画像データが色情報をもったランレングスデータで表現されているため、色毎の領域抽出処理を行ない、写真と同じ領域で抽出された同一色領域の大きさや形状や画素の分布といった特徴量を抽出してもよい。その結果、色に偏りが無い場合、写真と判定し、そのまま修正領域データ１５４として出力する（ステップＳＴ４０５）。偏りがある場合、線図形と判定し、写真属性から線図形属性に修正した修正領域データ１５４を出力する（ステップＳＴ４０６）。
【００６２】
以上の処理を全ての写真領域に行なうことで、識別精度の向上を図ることが可能となる。
【００６３】
図８は上記した本発明の第１実施形態に於ける画像処理装置に入力される一例としての画像データの例である。ここでは画像データ５０１として、線図形部分５０２と、文章部分５０３が描かれているものを示してある。この例を用いて画素属性識別部１０４に於ける実際の処理過程を説明する。
【００６４】
上記図８に示す画像データ５０１を画像入力部１０１より取り込み、領域識別部１０２によって領域として抽出した線図形部分５０２は、複雑なグラフであるために、写真領域と誤識別され、文章部分５０３は文字領域として識別される。その結果は、例えば、線図形領域、は写真領域と誤識別され、図９の５０２ａの如く、文字領域は５０３ａの如く出力される。
【００６５】
さて、写真領域と誤識別された線図形領域５０２ａは、画素属性識別部１０４によって写真であるか否か再判定される。この際は、色分類部１０３で生成された色分布画像データによって、写真領域と誤識別された線図形領域５０２ａは色が少ないと判定されたとする。その後、色分布画像データから同一色領域の抽出を行ない、その大きさや形状や画素の分布状態から、色の分布に偏りがあると判定される。したがって、画素属性識別部１０４の出力としては、修正領域データとして、写真領域を線図形領域と変更し、出力することとなる。
【００６６】
この例のように、本発明による画像処理装置によると複雑な線図形領域を写真領域として誤識別した場合であっても、色の分布状態から、線図形として正しく識別することが可能となる。
【００６７】
以上が、画素属性識別部１０４の処理の詳細である
また、別の形態では、色分類部１０３で生成される色分布画像データにしたがって、出力を制御することも可能である。この機能を実現した画像処理装置の一例を本発明の第２実施形態として説明する。
【００６８】
以下に本発明の第２実施形態に於ける、より具体的な処理の一例について、図面を参照して説明する。
【００６９】
図１０は本発明の第２実施形態に於ける画像処理装置の一構成例を示すブロック図である。
【００７０】
図１０に於いて、１０１は画像入力部、１０２は領域識別部、１０３は色分類部、１０４は画素属性識別部、１０５は画像出力部、１０６は画像出力制御部である。
【００７１】
画像入力部１０１、領域識別部１０２、色分類部１０３、および画素属性識別部１０４については、上述した第１実施形態に於いて、図１を用いて説明した構成要素と同様のものであり、ここではその説明を省略する。
【００７２】
画像出力制御部１０６は、色分類部１０３で生成される色分布画像データにしたがって、画像出力方法の変更を行ない、その情報を画像出力部１０５へ渡す。
【００７３】
図１１は本発明の画像処理装置に入力される一例としての画像データの例である。ここでは画像データ６０１として、写真部分６０２と、文章部分６０３と、閉曲線６０４が描かれているものを示している。文章部分６０３は黒で、閉曲線６０４は赤で描かれているとする。この例を用いて画像出力制御部１０６に於ける処理の一例を説明する。
【００７４】
画像入力部１０１より取り込まれた画像データ６０１は、領域識別部１０２によって領域抽出される。ここでは、図１２に於ける写真領域６０２ａと文字領域６０３ａと赤の平曲線６０４ａとに識別される。
【００７５】
画像出力制御部１０６は、抽出された領域の状態と色が対応付けられたルールによって構成される。例えば、「赤の閉曲線で囲まれた領域内に存在する文字の色を赤で置き換える」というルールが記述されているならば、色分類部１０３で生成される色分布画像データを参照することによって、閉曲線６０４ａ内に存在する文字領域６０３aの文字を「黒」から「赤」に置き換えることが可能となる。上述した従来技術では、閉曲線６０４ａ内の全て黒が赤に置き換えられてしまうが、本発明に於いては、画素属性とその色を対応付けることが可能であるため、このような高度な編集も可能となる。
【００７６】
また、別の形態では、画像出力部１０５が画像ファイリングであっても効果がある。
【００７７】
以下、本発明の第２実施形態に於ける、より具体的な処理の一例について、図面を参照して説明する。
【００７８】
図１３は本発明の第２実施形態に於ける画像処理装置に入力される一例としての画像データの例である。ここでは画像データ７０１として、線図形部分７０２と、文章部分７０３が描かれており、更に文字７０４と文字７０５は赤で描かれているとする。この例を用いて、上記画像出力制御部１０６に於ける処理の一例を説明する。
【００７９】
画像入力部１０１より取り込まれた上記画像データ７０１は、領域識別部１０２によって領域抽出される。ここでは、図１４に於ける線図形領域７０２ａと文字領域７０３ａとに識別される。
【００８０】
通常、文書内に於ける赤い文字は、その文書に於いて重要な意味を成すことが多いため、ファイリング時のキーワードとして利用することが考えられる。
【００８１】
画像出力制御部１０６で「赤の文字が存在する場合、その文字領域に対して文字認識処理を行なわせ、キーワードとして登録する」というルールを記述しておくと、上述した色分類部１０３で生成される色分布画像データを参照することによって、文字領域７０３ａ中に赤で描かれている文字７０４ａと７０５ａを自動的に抽出して、キーワード登録することも可能となる。
【００８２】
このように本発明の実施形態に於いては、画像データを画像入力部１０１により取り込み、この画像データ（カラー画像）を領域識別部１０２に於いて画素値の濃度差や彩度などの状態により複数の画像データに分離し、物理的あるいは論理的に連結しているものを一つの領域として抽出した後、個々の領域の画像上の位置、大きさ、形状、構造、濃度分布等の特徴量を計測し、更に、その計測結果を予め定められたルールに基づいて文書構成要素として識別し、色分類部１０３に於いて画素値の色相や彩度などの状態によって、幾つかの色に分類し、その結果を色分布画像データとして保持し、画素属性識別部１０４に於いて領域識別データと色分布画像データから文書内の各画素の属性を決定し、画像出力部１０５によって、構造的な画素属性を出力している。このような処理機能をもつことで、文書要素の識別精度を向上させることが可能となり、また色の違いを利用した、より高度な編集処理を行なうことが可能となる。
【００８３】
【発明の効果】
以上詳記したように、本発明によれば、画像データとして読み込まれた画像の各画素を、前記画像の画素およびその近傍画素の濃度情報に応じて複数の画像に分離し、当該各画像から領域を抽出し、抽出した領域の特徴量により分離された画像毎の属性を識別する画像処理に於いて、例えば写真領域、線図形領域等を弁別するための有効な処理が期待できる。
【００８４】
更に、本発明によれば、文書をハードコピーしたり、イメージデータに変換して保存しようとする場合、入力されたカラー画像に対して、画素値の濃度差や彩度などの状態によって複数の画像データに分離し、物理的あるいは論理的に連結しているものを一つの領域として抽出した後、個々の領域の画像上の位置、大きさ、形状、構造、濃度分布等の特徴量を計測し、その計測結果を予め定められたルールに基づいて文書構成要素として識別し、画素値の色相や彩度などの状態によって、幾つかの色に分類して、その結果を色分布画像データとして保持し、領域識別結果と色分布画像データの比較により、文書要素の識別精度を向上させることが可能となる。その結果、従来発生していた写真領域や線図形領域の誤識別の軽減につながり、例えばハードコピーをとる場合、写真領域に対してはローパスフィルタ処理を行って滑らかな階調を再現するためにローパスフィルタを施すことによって、ざらつき感が無くなり、奇麗な出力となる。また、線図形領域ではハイパスフィルタ処理を施すことによって、文字のエッジ部分が強調され、くっきりとした出力にすることが可能となる。また、色によって画像編集の指示を行なったり、ファイリングの際に色付きの文字をキーワードとして自動登録を行なったりすることも可能となる。
【図面の簡単な説明】
【図１】本発明の第１実施形態に於ける画像処理装置の要部の構成要素を示すブロック図。
【図２】上記実施形態に於ける画像処理手順の一例を示すフローチャート。
【図３】上記実施形態に於ける色分類部の処理手順の一例を示すフローチャート。
【図４】上記実施形態に於ける色分類部の処理動作を説明するための画素変換のカラーモデルを示す図。し、図５はそのモデルの断面を表している。
【図５】上記図４に示すカラーモデルの断面を示す図。
【図６】上記実施形態に於ける色分布画像生成処理の動作説明図。
【図７】上記実施形態に於ける画素属性識別部の処理手順を示すフローチャート
【図８】上記実施形態の動作を説明するための入力画像データの一例を示す図。
【図９】上記図８に示す画像データの領域抽出例を示す図。
【図１０】本発明の第２実施形態に於ける画像処理装置の要部の構成要素を示すブロック図。
【図１１】上記実施形態の動作を説明するための入力画像データの一例を示す図。
【図１２】上記図１１に示す画像データの領域抽出例を示す図。
【図１３】上記実施形態の動作を説明するための入力画像データの一例を示す図。
【図１４】上記図１３に示す画像データの領域抽出例を示す図。
【符号の説明】
１０１…は画像入力部
１０２…領域識別部
１０３…色分類部
１０４…画素属性識別部
１０５…画像出力部
１０６…画像出力制御部
５０１…画像データ
５０２…領域として抽出した線図形部分
５０３…領域として抽出した文章部分
５０２ａ…写真領域と誤識別された線図形領域
５０３ａ…文章領域と誤識別された線図形領域
６０１…画像データ
６０２…領域として抽出した写真部分
６０３…領域として抽出した文章部分
６０４…領域として抽出した閉曲線部分
６０２ａ…識別された写真領域
６０３ａ…識別された文字領域
６０４ａ…識別された平曲線領域
７０１…画像データ
７０２…領域として抽出した線図形部分
７０３…領域として抽出した文章部分
７０４…領域として抽出した文字部分
７０４，７０５…文字領域に赤で描かれた文字
７０２ａ…識別された線図形領域
７０３ａ…識別された文字領域
７０４ａ，７０５ａ…自動抽出された赤で描かれている文字[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image processing apparatus and an image processing method for converting an image of a document or the like by filing or copying using document components and color information.
[0002]
[Prior art]
Technology used in the field of image capture in copying machines, personal computers, etc., as a technology for capturing document images as image data and outputting the images as hard copies or saving them as image files And technologies used for filing system devices and databases.
[0003]
In this type of apparatus that handles image data, when the captured image is output as a hard copy, image processing may be performed to make it easy for the user to view as a document image. For example, when a character is included in a document image, the edge portion of the character is emphasized by performing a high-pass filter process, and the character becomes clear and easy to read. Further, in the case of a photograph, by applying a low-pass filter to reproduce the smooth gradation of the photograph, the feeling of roughness is eliminated and a beautiful output is obtained. It is also possible to switch the processing depending on whether the photograph is a halftone dot photograph composed of halftone dots or a silver salt photograph.
[0004]
On the other hand, when capturing as image data, the amount of data of color images and grayscale images is very large, and therefore data compression is often performed to reduce the amount of data. In such a case, it is possible to reduce the amount of data by binarizing characters, and the amount of data can be reduced while preserving the beauty of the photograph by applying compression processing that preserves gradation. Can be compressed.
[0005]
As described above, changing the processing method according to the type of document image is very effective both when outputting as a hard copy and when saving as an image file.
[0006]
A document image is usually constituted by areas of text, diagrams, photographs, etc., each independently or overlapping. Therefore, when image processing is performed, it is necessary to detect what region and how the document image is configured.
[0007]
Conventionally, a method using a density value of input image data has been used to identify elements constituting a document image such as characters, halftone dots, and photographs. A method of separating a photo area and a non-photo area using a difference between the maximum value and the minimum value in a block of about 2 × 2 to 4 × 4 pixels, detecting an edge pixel in the block, and the edge pixel is a character A method for determining whether or not it is part of a pattern by pattern matching, a method for determining by performing edge detection and peak pixel detection by pattern matching on a ternary image, and summing the amount of edges at each pixel in the block A method of calculating and separating. What is common to these methods is that processing is performed for each block of several pixels in order to realize high-speed processing.
[0008]
Therefore, depending on the change of the pixel value existing in the block, there is a possibility that it is mistakenly identified as a different element between adjacent blocks in the same component, and as a result, different images in the same area. Processing is performed.
[0009]
As a method for solving this, there is a method for extracting the structure of a document image using layout analysis, as described in Japanese Patent Application No. 8-34702. In this method, after performing the binarization process, the connectivity of the pixels is examined, the connected pixels are extracted as regions, and are identified using the feature amount such as the position and size. is there. Further, as described in Japanese Patent Application No. 10-191286, there is a method in which an input image is separated into a plurality of binary images and layout analysis is performed on each binary image data. . In this method, images are separated based on density values. Therefore, by comparing the results of processing the images, it is possible to accurately extract and identify document elements, which is difficult with normal binarization. . In addition, since color images and grayscale images are expressed by a plurality of binary images, processing time and memory capacity are reduced.
[0010]
However, in these methods, although it was possible to give color attributes based on differences in saturation such as color, gray, and black for each document element, it was not known what color was used for drawing.
[0011]
For example, in a color image, a complicated line figure such as a graph may have a feature amount similar to that of a photograph in the region, and may be erroneously identified as a photograph. Conversely, a thin photographic area may have a feature quantity similar to that of a line figure, and thus may be erroneously identified as a line figure. As a result of erroneous identification, there is a possibility that a high-pass filter process is performed on a photographic area to produce a rough image, and conversely, a line figure is subjected to a low-pass filter process to produce an image with blurred characters and lines. Usually, a color photograph is expressed by multiple colors, and the color distribution is often scattered. A line figure is represented by a limited color, and the color distribution is often like a lump. Therefore, it is possible to prevent erroneous identification by examining the color distribution. However, it is difficult to examine the color distribution only with color attributes.
[0012]
In general, when a document has red characters, it is often a portion that should be noted more than other portions. Since such characters are assumed to be keywords when filing, it is important to distinguish them from other characters when considering more advanced document processing. However, in this case as well, it is difficult to distinguish red characters from other colors only with color attributes.
[0013]
Furthermore, when outputting as a hard copy, if the colors used in advance are known, when the color conversion table is constructed, the unused colors are excluded and the resolution of the colors used is reversed. As a result, it is possible to obtain a beautiful output, but it is difficult to construct such an adaptive color table because the color distribution is unknown.
[0014]
As described above, in the method described in Japanese Patent Application No. 10-191286, the color distribution indicating the color in which the document element is drawn is not known. Therefore, the feature in the color image is extracted. It was difficult.
[0015]
[Problems to be solved by the invention]
As described above, conventionally, each pixel of an image read as image data is separated into a plurality of images according to density information of the pixel of the image and its neighboring pixels, and an area is extracted from each image and extracted. In the image processing for identifying the attribute of each image separated by the feature amount of the area, there is no effective processing means for discriminating, for example, a photographic area, a line figure area, and the like.
[0016]
The present invention has been made in view of the above circumstances, and each pixel of an image read as image data is separated into a plurality of images according to density information of the pixels of the image and its neighboring pixels, and an area from each image is obtained. And an image processing apparatus that can be expected to be effective for discriminating, for example, a photograph area, a line figure area, and the like in image processing for identifying attributes for each image separated by the feature amount of the extracted area, and An object is to provide an image processing method.
[0017]
[Means for Solving the Problems]
The present invention separates each pixel of an image read as image data into a plurality of images according to density information of the pixel of the image and its neighboring pixels, extracts a region from each image, and extracts the extracted region In an image processing method for identifying an attribute for each image separated by a feature amount, a shape of a color cluster of a plurality of colors or a specific color in the extracted region is recognized, and information on the recognized color cluster state is obtained. The attribute of the image is identified by an identification condition.
[0018]
In addition, the image processing apparatus according to the present invention separates each pixel of an image read as image data into a plurality of images according to density information of the pixel of the image and its neighboring pixels, and a region is divided from each image. Region identification means for identifying the attribute for each image extracted and separated by the feature quantity of the extracted region, and each pixel of the image read as the image data by the color information of the pixel of the image and its neighboring pixels Each pixel of the image read as the image data based on the color classification means for classifying into several kinds of colors, the attribute of the image acquired by the area identification means, and the information of the color acquired by the color classification means A pixel attribute identifying means for identifying the attribute of the image, and an image output means for converting the attribute identified by the pixel attribute identifying means into an image and outputting the image.
[0019]
In addition, the image processing apparatus according to the present invention separates each pixel of an image read as image data into a plurality of images according to density information of the pixel of the image and its neighboring pixels, and a region is divided from each image. Region identification means for identifying the attribute for each image extracted and separated by the feature quantity of the extracted region, and each pixel of the image read as the image data by the color information of the pixel of the image and its neighboring pixels Each pixel of the image read as the image data based on the color classification means for classifying into several kinds of colors, the attribute of the image acquired by the area identification means, and the information of the color acquired by the color classification means A pixel attribute identifying unit that determines an attribute of the image, an image output control unit that changes an image generation method for the specific attribute acquired by the pixel attribute identifying unit, and a pixel attribute identifying unit. Characterized by comprising an image output means for converting the attributes in the image.
[0020]
In addition, the image processing apparatus according to the present invention separates each pixel of an image read as image data into a plurality of images according to density information of the pixel of the image and its neighboring pixels, and a region is divided from each image. Area identification means for identifying attributes for each image extracted and separated by the feature quantity of the extracted area, and color distribution information in the area from the color information of each pixel in the area extracted by the area identification means The color classification means for acquiring the image attribute, the attribute of the image acquired by the area identification means, and the attribute of each pixel of the image read as the image data based on each information of the color acquired by the color classification means It comprises a pixel attribute identifying means for identifying, and an image output means for converting the attribute identified by the pixel attribute identifying means into an image and outputting it.
[0021]
By providing the above-described functions of the present invention, it is possible to reduce the erroneous identification of a photographic area or a line figure area that has occurred in the past. For example, when taking a hard copy, a low-pass filter process is performed on the photographic area for smoothness. By applying a low-pass filter to reproduce a smooth gradation, the feeling of roughness is eliminated, resulting in a beautiful output, and by applying high-pass filter processing to the line figure area, the edge portion of the character is emphasized, It becomes possible to obtain a clear output. In addition, it is possible to instruct image editing by color, or to perform automatic registration using colored characters as keywords during filing.
[0022]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0023]
FIG. 1 is a block diagram showing a configuration of an apparatus system according to the first embodiment of the present invention.
[0024]
In the present invention, a document image is captured by an image input unit, and is divided into a plurality of image data according to a state such as a pixel value density difference or saturation in a region identification unit, and is physically or logically connected. Are extracted as a single area, and then the features, such as the position, size, shape, structure, density distribution, etc. on the image of each area are measured, and the measurement results are based on predetermined rules. In the color classification unit, it is classified into several colors according to the state of the pixel value such as hue and saturation, the result is held as color distribution image data, and further extracted from a plurality of image data By determining the attribute of each pixel of the image from the area information and the color distribution image data, the attribute for each area in the document image can be detected. In addition, the image output control unit can cause the pixel having a specific attribute to be edited based on the color distribution image data.
[0025]
Hereinafter, more specific embodiments of the present invention will be described with reference to the drawings.
[0026]
FIG. 1 is a block diagram showing components of a main part of the image processing apparatus according to the first embodiment of the present invention.
[0027]
In FIG. 1, 101 is an image input unit, 102 is a region identification unit, 103 is a color classification unit, 104 is a pixel attribute identification unit, and 105 is an image output unit.
[0028]
The image input unit 101 is a device that inputs image data, and is a device that captures what is drawn from a document such as a document by an image input device such as an image scanner that is a device that reads a document and converts it into image data. is there. The image input unit 101 may be configured by a reading device such as an image scanner, or may be a device that captures a stored image image.
[0029]
The area identification unit 102 extracts document elements as areas from the image data input by the image input unit 101. First, the image data input in the image input unit 101 is separated into a plurality of binary image data depending on the density difference and saturation of peripheral pixels, and characters and figures are physically displayed from each image. By dividing into areas that are logically or logically connected, and measuring the amount of features such as position, size, shape, structure, density distribution, etc. Identify the degree. Examples of document element types include characters, photographs, diagrams, tables, and halftone dots. As a specific method of separating a plurality of binary image data, there is a known method, and for example, it may be realized by a method disclosed in Japanese Patent Application No. 10-053317. In this case, seven binary separated image data of a character image, a halftone image, a background image, a halftone image, a color image, a gray image, and a black image are generated. In addition, as a specific method for extracting and identifying document elements, there are already known methods, and for example, it may be realized by a method disclosed in Japanese Patent Application No. 8-34702. The area extraction unit 102 integrates feature quantities not only from single image data but also from multiple image data according to a rule, and determines the area attribute of the input image data. For example, when an area is extracted at the same position from both a character image and a halftone image, the type of the area and the size of the area are determined. As a specific example, when a photographic region exists on a character image and a halftone pixel exists in the halftone image at the same position, it is determined as a silver salt photographic region. As a specific method for determining region attributes from a plurality of image data, there are already known methods, and for example, it may be realized by the method disclosed in Japanese Patent Application No. 10-053317.
[0030]
Although the details will be described later, the color classification unit 103 performs several colors (for several types of colors on the basis of the state of hue and saturation of peripheral pixels) for the image data input by the image input unit 101. And the result is generated as color classified image data.
[0031]
Although details will be described later, the pixel attribute identification unit 104 is input using the type and importance of the region of the multiple image data identified by the region identification unit 102 and the color classification image data classified by the color separation unit 103. The attribute for each pixel of the image data is determined.
[0032]
The image output unit 105 performs output according to the region information determined by the pixel attribute identification unit 104. As an output form, for example, attribute region information may be converted into an image. As a specific method for converting attribute area information into an image, there is a known method, and for example, it may be realized by a method disclosed in Japanese Patent Application No. 10-053317.
[0033]
FIG. 2 is a flowchart showing an example of a document processing procedure of the image processing apparatus according to the first embodiment of the present invention shown in FIG.
[0034]
The processing operation in the first embodiment of the present invention will be described with reference to the respective drawings.
[0035]
First, a document image is captured by the image input unit 101 (step ST201). That is, an image input unit 101 converts an image input device such as a scanner that reads an image from a document or inputs image file data such as a filing system into image data.
[0036]
The area extraction unit 102 reads the image data output from the image input unit 101 for one line or several lines, and a plurality of binary image data depending on the state of each pixel such as density difference and saturation of surrounding pixels. In addition, it is divided and extracted for each area where characters or figures are physically or logically connected, and the position, size, shape, structure, density distribution, etc. of the area are measured. Then, the type and importance of the area are identified, and as a result, document elements such as characters, photographs, diagrams, tables, and dot areas are extracted (step ST202). This process is repeated until the process is completed for all pixels of the input image (step ST203).
[0037]
The color classification unit 103 reads the image data output from the image input unit 101 for one line or several lines, and determines the color of each pixel depending on the state of each pixel such as hue and saturation of surrounding pixels. And color distribution image data is generated. At this time, for data compression, the minute color distribution is replaced with the surrounding large color distribution from the color distribution of the peripheral pixels (step ST204). This process is repeated until the process is completed for all pixels of the input image (step ST205).
[0038]
After region identification and color classification are completed for all pixels, the pixel attribute determination unit 104 determines the final attribute of each pixel (step ST206). An example of the attribute determined here is a distinction between a photograph and a line figure, addition of color distribution information, and the like. Then, the above processing is repeated for all the pixels (step ST207).
[0039]
Thereafter, the image output unit 105 outputs an output image obtained by converting each pixel attribute as image data (step ST208).
[0040]
The above is the rough processing operation of the image processing apparatus according to the first embodiment of the present invention.
[0041]
Next, details of the processing of individual elements in the above embodiment will be described.
[0042]
FIG. 3 is a flowchart showing an example of the processing procedure of the color classification unit 103 provided in the image processing apparatus according to the first embodiment of the present invention, and shows the processing procedure performed in step ST204 shown in FIG. It is a flowchart.
[0043]
The color classification unit 103 classifies the input image into several colors based on the characteristics of the hue, saturation, and luminance value of the input image data. Here, the number of colors is determined by the content to be processed by the pixel attribute identification unit 104. If it is necessary to check the color strictly, the number of classifications is increased. Conversely, if the rough color distribution is known, the number of classifications is decreased. In general, when the number of classifications is increased, the amount of data increases, causing problems in terms of cost. In the present invention, the case where the latter rough color distribution is known will be described.
[0044]
In the example below, the input pixels are red, yellow, green, cyan, blue, magenta, white, gray, An example of classification into nine colors of black (black) is shown.
[0045]
The image data 151 input from the image input unit 101 includes color information expressed in RGB. As an example, the RGB image data is converted into a luminance value Y by YUV linear conversion expressed by the following equation (step ST301).
[0046]
_{Y = 0.299R + 0.587G + 0.114B}
Here, R, G, and B are the RGB values of the pixels in the input image data. Further, the image data expressed in the same RGB is converted into saturation S using the conversion formula shown in Table 1 as an example.
[0047]
[Table 1]

[0048]
FIG. 4 shows a color model of pixel conversion shown in the above example, and FIG. 5 shows a cross section of the model. The small circular area in the center of FIG. 5 is an area that becomes one of white, gray, and black.
[0049]
Depending on the value of H, it is classified into one of red, yellow, green, cyan, blue, and magenta.
[0050]
As a result of the above processing, each pixel is red, yellow, green, cyan, blue, magenta, white, gray, black (black)
[0051]
These conversion processes are performed for one line of the input image data, and the result is stored in the memory.
[0052]
Next, a rougher color distribution is determined from the value of H (step ST302). Here, the color distribution is determined while reducing the resolution in the x direction from the color distribution data of the current row and the color distribution data of the previous row.
[0053]
_xWhen the resolution of the direction is reduced to 1 / N and the pixel position for determining the color distribution is (i, j), the color distribution of the 3N pixels before and after the i-1 row and the 3N pixels before and after the i row is examined. Count is performed for each of red, yellow, green, cyan, blue, magenta, white, gray, and black. As a result, the color having the largest count number is determined as the color distribution of the pixel.
[0054]
FIG. 6 shows an example in which N = 3. In the figure, r represents red, m represents magenta, g represents gray, and b represents black. When each color is counted in the corresponding pixel, r = 8, g = 5, m = 3, and b = 2. From this, the color distribution can be determined to be red. Through the above processing, it is possible to determine a color distribution that shows a rough color distribution state with a reduced resolution.
[0055]
Then, color distribution image data is generated from the determined color distribution (step ST303). This can be further compressed by describing the data with the expression “color identifier + number of consecutive pixels” like run-length data. As a result, color distribution image data 152 is output.
[0056]
As described above, the color distribution image data is generated as the output of the color classification unit 103.
[0057]
Next, pixel attribute identification processing will be described.
FIG. 7 is a flowchart showing a pixel attribute identification process procedure of the pixel attribute identification unit 104 provided in the image processing apparatus according to the first embodiment of the present invention, and shows details of the process performed in step ST206 of FIG. It is a flowchart.
[0058]
The pixel attribute identification process is a process of confirming and correcting the area identification result by comparing the results of area identification and color classification. Hereinafter, the pixel attribute identification processing operation will be described with reference to the flowchart shown in FIG.
[0059]
The area identifying unit 102 extracts document elements, identifies them, and outputs the results. The color classification unit 103 classifies several kinds of colors and outputs the result as color distribution image data. The pixel attribute identification unit 104 re-determines the region identification result from the region identification result and the information of the color distribution image data at the position corresponding to the region identification result. In the following example, a re-determination method for an area identified as a color photograph will be described. In addition, the photograph here points out a color photograph.
[0060]
First, it is checked whether or not the region identification data 153 has a photograph attribute (step ST401). If it does not have a photograph attribute, for example, if it is a text attribute, it is output as it is as the correction area data 154.
[0061]
If the area identification data 153 has a photograph attribute in step ST401, the number of colors in the color distribution information of the corresponding area is checked from the color distribution image data 152 (step ST402). If it is a photograph, it is often expressed in various colors. In such a region, a number of colors are detected as a result of the color classification. On the contrary, if the result of color classification is close to a single color, it can be regarded as a complicated line figure. If there are many colors, it is determined as a photograph and output as correction area data 154 as it is (step ST403). On the other hand, if it is close to a single color, the color distribution bias in the corresponding area is next examined from the color distribution image data 152 (step ST404). If it is a photograph, the color is often scattered throughout the area. Conversely, if it is a line figure, since a single color is often assigned to an object, it is expected that the color distribution will be biased. As a method for examining the color deviation, since the color distribution image data is represented by run-length data having color information, an area extraction process is performed for each color, and the same color area extracted in the same area as the photograph is extracted. Features such as size, shape, and pixel distribution may be extracted. As a result, if there is no color deviation, it is determined as a photograph and output as it is as the correction area data 154 (step ST405). If there is a bias, it is determined as a line figure, and the corrected area data 154 corrected from the photograph attribute to the line figure attribute is output (step ST406).
[0062]
It is possible to improve the identification accuracy by performing the above processing on all the photo areas.
[0063]
FIG. 8 shows an example of image data input as an example to the image processing apparatus according to the first embodiment of the present invention. Here, as the image data 501, a line figure portion 502 and a sentence portion 503 are depicted. The actual process in the pixel attribute identification unit 104 will be described using this example.
[0064]
The line figure portion 502 obtained by taking the image data 501 shown in FIG. 8 from the image input unit 101 and extracted as a region by the region identifying unit 102 is a complicated graph, and thus is misidentified as a photographic region. Identified as a character area. As a result, for example, a line figure area is mistakenly identified as a photographic area, and a character area is output as 503a as in 502a of FIG.
[0065]
Now, the line attribute area 502a erroneously identified as a photograph area is re-determined by the pixel attribute identification unit 104 as to whether or not it is a photograph. In this case, it is assumed that the line figure region 502a erroneously identified as a photographic region is determined to have few colors based on the color distribution image data generated by the color classification unit 103. Thereafter, the same color region is extracted from the color distribution image data, and it is determined that the color distribution is biased from the size, shape, and pixel distribution state. Therefore, as the output of the pixel attribute identification unit 104, the photograph area is changed to the line figure area as the correction area data and is output.
[0066]
As in this example, according to the image processing apparatus of the present invention, even when a complicated line figure area is erroneously identified as a photographic area, it can be correctly identified as a line figure from the color distribution state.
[0067]
The above is the details of the processing of the pixel attribute identification unit 104.
In another embodiment, the output can be controlled according to the color distribution image data generated by the color classification unit 103. An example of an image processing apparatus that realizes this function will be described as a second embodiment of the present invention.
[0068]
Hereinafter, an example of more specific processing in the second embodiment of the present invention will be described with reference to the drawings.
[0069]
FIG. 10 is a block diagram showing an example of the configuration of the image processing apparatus according to the second embodiment of the present invention.
[0070]
In FIG. 10, 101 is an image input unit, 102 is a region identification unit, 103 is a color classification unit, 104 is a pixel attribute identification unit, 105 is an image output unit, and 106 is an image output control unit.
[0071]
The image input unit 101, the region identification unit 102, the color classification unit 103, and the pixel attribute identification unit 104 are the same as the components described with reference to FIG. 1 in the first embodiment described above. The description is omitted here.
[0072]
The image output control unit 106 changes the image output method according to the color distribution image data generated by the color classification unit 103 and passes the information to the image output unit 105.
[0073]
FIG. 11 shows an example of image data input to the image processing apparatus of the present invention. Here, as the image data 601, a photograph portion 602, a sentence portion 603, and a closed curve 604 are illustrated. It is assumed that the sentence portion 603 is drawn in black and the closed curve 604 is drawn in red. An example of processing in the image output control unit 106 will be described using this example.
[0074]
The image data 601 captured from the image input unit 101 is subjected to region extraction by the region identification unit 102. Here, it is distinguished into a photo area 602a, a character area 603a, and a red flat curve 604a in FIG.
[0075]
The image output control unit 106 is configured by a rule in which the state and color of the extracted area are associated with each other. For example, if the rule “replace character color existing in the region surrounded by the red closed curve with red” is described, the color distribution image data generated by the color classification unit 103 is referred to. Thus, it is possible to replace the characters in the character region 603a existing in the closed curve 604a from “black” to “red”. In the above-described prior art, all black in the closed curve 604a is replaced with red. However, in the present invention, pixel attributes can be associated with their colors, and such advanced editing is also possible. It becomes.
[0076]
In another form, the image output unit 105 is effective even when it is image filing.
[0077]
Hereinafter, an example of more specific processing in the second embodiment of the present invention will be described with reference to the drawings.
[0078]
FIG. 13 shows an example of image data that is input to the image processing apparatus according to the second embodiment of the present invention. Here, it is assumed that a line figure portion 702 and a sentence portion 703 are drawn as the image data 701, and that the characters 704 and 705 are drawn in red. An example of processing in the image output control unit 106 will be described using this example.
[0079]
The image data 701 fetched from the image input unit 101 is subjected to region extraction by the region identification unit 102. Here, a line figure area 702a and a character area 703a in FIG. 14 are identified.
[0080]
Usually, red characters in a document often have an important meaning in the document, so it may be used as a keyword for filing.
[0081]
If the rule that “when a red character exists, the character area is subjected to character recognition processing and registered as a keyword” is described in the image output control unit 106, the color classification unit 103 generates the rule. By referring to the color distribution image data, the characters 704a and 705a drawn in red in the character region 703a can be automatically extracted and registered as keywords.
[0082]
As described above, in the embodiment of the present invention, image data is captured by the image input unit 101, and the image data (color image) is captured by the region identification unit 102 according to the state of density difference or saturation of the pixel value. After separating a plurality of image data and physically or logically connecting them as one area, the feature values such as the position, size, shape, structure, density distribution, etc. of each area on the image In addition, the measurement result is identified as a document component based on a predetermined rule, and is classified into several colors by the color classification unit 103 according to the state of the pixel value such as hue and saturation. The result is stored as color distribution image data, and the pixel attribute identification unit 104 determines the attribute of each pixel in the document from the area identification data and the color distribution image data. Picture And it outputs the attribute. By having such a processing function, it is possible to improve the identification accuracy of the document element, and it is possible to perform a more advanced editing process using the difference in color.
[0083]
【The invention's effect】
As described above in detail, according to the present invention, each pixel of an image read as image data is separated into a plurality of images according to the density information of the pixel of the image and its neighboring pixels, and from each of the images. In the image processing for extracting the region and identifying the attribute for each image separated by the feature amount of the extracted region, it is possible to expect an effective process for discriminating, for example, a photograph region, a line figure region, and the like.
[0084]
Furthermore, according to the present invention, when a document is to be hard-copied or converted to image data and stored, a plurality of input color images can be selected depending on states such as pixel value density differences and saturation. After separating the image data and physically or logically connecting them as one area, measure the amount of features such as the position, size, shape, structure, and density distribution of each area on the image. The measurement result is identified as a document component based on a predetermined rule, and is classified into several colors according to the state of the pixel value such as hue and saturation, and the result is used as color distribution image data. The identification accuracy of the document element can be improved by holding and comparing the region identification result and the color distribution image data. As a result, it has led to the reduction of misidentification of photographic areas and line figure areas that have occurred in the past.For example, when taking a hard copy, low-pass filter processing is performed on the photographic area to reproduce smooth gradations. By applying a low-pass filter, the feeling of roughness is eliminated and a beautiful output is obtained. In addition, by applying high-pass filter processing in the line figure region, the edge portion of the character is emphasized, and a clear output can be obtained. In addition, it is possible to instruct image editing by color, or to perform automatic registration using colored characters as keywords during filing.
[Brief description of the drawings]
FIG. 1 is a block diagram showing components of a main part of an image processing apparatus according to a first embodiment of the present invention.
FIG. 2 is a flowchart showing an example of an image processing procedure in the embodiment.
FIG. 3 is a flowchart showing an example of a processing procedure of a color classification unit in the embodiment.
FIG. 4 is a diagram showing a color model of pixel conversion for explaining a processing operation of a color classification unit in the embodiment. FIG. 5 shows a cross section of the model.
FIG. 5 is a view showing a cross section of the color model shown in FIG. 4;
FIG. 6 is an operation explanatory diagram of color distribution image generation processing in the embodiment.
FIG. 7 is a flowchart showing a processing procedure of a pixel attribute identification unit in the embodiment.
FIG. 8 is a diagram showing an example of input image data for explaining the operation of the embodiment.
9 is a diagram showing an example of region extraction of the image data shown in FIG.
FIG. 10 is a block diagram showing components of a main part of an image processing apparatus according to a second embodiment of the present invention.
FIG. 11 is a diagram showing an example of input image data for explaining the operation of the embodiment.
12 is a diagram showing an example of region extraction of the image data shown in FIG.
FIG. 13 is a diagram showing an example of input image data for explaining the operation of the embodiment.
14 is a diagram showing an example of region extraction of the image data shown in FIG.
[Explanation of symbols]
101 is an image input unit
102 ... Area identification unit
103 ... Color classification part
104: Pixel attribute identification unit
105: Image output unit
106: Image output control unit
501: Image data
502 ... Line figure portion extracted as a region
503 ... sentence part extracted as a region
502a ... Line figure area misidentified as a photograph area
503a: Line figure area misidentified as text area
601: Image data
602 ... Photo part extracted as an area
603 ... sentence part extracted as a region
604 ... Closed curve portion extracted as a region
602a ... Photo area identified
603a: Identified character area
604a ... identified flat curve region
701: Image data
702 ... Line figure portion extracted as a region
703 ... sentence part extracted as a region
704 ... Character portion extracted as a region
704, 705 ... Characters drawn in red in the character area
702a: identified line figure area
703a: identified character area
704a, 705a ... automatically drawn characters drawn in red

Claims (4)

Into a plurality of images in accordance with each pixel and the density information of the neighboring pixels of the read input image as image data, from the respective image extracting area, extracted area for each separated image by the feature of Area identifying means for identifying a plurality of area attributes including photo attributes and text attributes ;
Color classification means for classifying each pixel of the input image into several kinds of colors according to color information of the pixel of the input image and its neighboring pixels;
If the area attribute of the image identified by the area identifying means is the photograph attribute, based on each information of the color which has been classified by the color classification means, whether the area attribute is truly photograph attribute re A pixel attribute identifying unit for identifying the region based on the number of colors of the region acquired from the color information classified by the color classification unit for the region identified as the photo attribute If the region attribute is identified as a true photo attribute and is not identified as a true photo attribute, the color classified by the color classifying unit is targeted for the region not identified as a true photo attribute. Based on this information , the presence / absence of bias in the distribution of the same color in the area is determined. If there is no bias in the distribution of the same color, the area attribute of the area is identified as a true photo attribute, and the same color If there is a distribution bias The image processing apparatus characterized by comprising a pixel attribute identifying means region attribute of the region to modify the area identification result by the area identifying means as a line graphic attribute.   2. The image processing apparatus according to claim 1, further comprising image output means for outputting an image in accordance with the region attribute identified by the pixel attribute identification means.   3. The image output control means according to claim 2, further comprising an image output control means for changing a color of an image having a specific attribute existing in an area whose attribute is identified by the pixel attribute identification means according to a predetermined rule. Image processing device. The image is separated into a plurality of images according to the density information of each pixel of the input image read as image data and its neighboring pixels, a region is extracted from each image, and each image separated by the extracted region feature amount is extracted. an area identification step of identifying a plurality of regions attributes including photo attributes and text attributes,
A classification step of classifying each pixel of the input image into several types of colors according to color information of the pixel of the input image and its neighboring pixels;
If the area attribute of the identified image in the region identification step is the photograph attribute, based on each information of the classified color, re-identification of re-identifying whether the photograph attribute is truly photograph attribute In the step, for the area identified as the photo attribute, based on the number of colors of the area acquired from the classified color information, the area attribute of the area is truly a photo attribute. Identifying the presence of the same color in the area based on the information on the classified color for an area that is not identified as a true photo attribute. Determining whether or not there is a bias in the distribution of the image, and if there is no bias in the distribution of the same color, the step of identifying that the region attribute of the region is truly a photo attribute, and the case where there is a bias in the distribution of the same color Is the region attribute of the region And re-identification step and a step of correcting the region identification result by the region identification step as a line graphics attribute
An image processing method comprising :

Images