TW200529650A - Video coding method and apparatus thereof - Google Patents

Abstract

A region-of-interest (ROI) video-coding method and apparatus based on fuzzy logic control for a video encoder is provided. Providing an image having a plurality of region-of-interest regions and a plurality of non-region-of-interest regions, the first step is to separate the region-of-interest regions and the non-region-of-interest regions from the image. Then by sending the region-of-interest regions from an input image to a fuzzy logic controller, in which the fuzzy logic controller performs fuzzy manipulations that enhance the quality of the region-of-interest regions, and therefore the region-of-interest quality of an output image will be improved. The method and apparatus are particularly useful in videophone and videoconferencing.

Description

200529650 V. Description of the invention α) Technical field to which the invention belongs The present invention relates to a method and device for improving image quality, and in particular, to a method and device for image coding of a target area for fuzzy logic control of an image encoder . 1 Prior Technology Recently, the demand for digital video communication using video conferences and video phones has increased. However, because the network transmission rate is limited, as a very low bit rate image coding for these applications, it is a kind of data rate that can be used to reduce the picture sequence without reducing its data rate. Important technology of quality. Most implementations of these standards are considered equal for each block (b 1 oc k). Although different blocks within the same image may be coded in different modes, no one block is more important than the other. This traditional model is not suitable for application in any region-in-interest (ROI) of a video sequence. In the H · 263 + standard, the distortion weight parameter and signal change at the macro-block (MB) layer are adjusted to control the quality of different regions. Blocks that correspond to the same focus area are more important than blocks in the background or other unnecessary areas. Although it sacrifices the quality of the background or non-emphasis area, it is possible to allocate more bandwidth to the area where the user cares. It is a good coding strategy for video sequences like video conferences. In addition to the high quality of ROI, some background information may also be ignored to improve encoding speed. For example, maximum bit transfer (MBT)

12429TWF.PTD Page 6 200529650 V. Description of the invention (2) The scene is always coded at the roughest quantization level ((1113111: 12 to 〇〇11616 乂 61). For this reason, traditionally, a region is used Region-based blurring algorithm 'to reduce the bit rate of the very low bit-rate video editing stone horse. Another method is to improve the quality of R 0 I and reduce the encoding background bit. Three fixed factors of each R 0 1 MB and non-ROI MB to improve the quality of R 0 I. The present invention can adaptively improve the quality of R 0I according to the fuzzy logic rate control, and is applicable to Immediate video conference. Fuzzy logic was first proposed by AZ Zadeh, who worked at Berkeley in 1965, and was determined after the natural population reached a three-point solution Model. The first point: solutions using different rules for the same problem. The second point: using more than one rule for the same problem at the same time. The third point: accepting a certain degree of uncertainty (imprecision). The solution is quite helpful. Obviously, the normal rate control rules used in different standard test models such as * TMN5, TMN8, etc. are in line with these two points. In each test model There are specific mathematical solutions' to determine the quantizati rl parameter of each MB, and accept the appropriate inaccuracies to estimate the bit rate of the next 0. It seems Fuzzy logic control is quite suitable for solving the rate control of image coding. Figure 1a is a block diagram of a conventional feedback control system. The controller in the figure is based on a mathematical model of processing, or mathematics.

12429TWF.PTD Page 7 200529650 V. Description of the invention (3) A fixed set of relationships determines how to proceed next. Figure 1b is a block diagram of a fuzzy logic control system 150. The fuzzy logic controller 150 uses a set of reaction rules formulated by an operator or system engineer with relevant experience as its operation guide (g u i d e). Referring to FIG. 1b, the quantizer 152 obtains data from the sensor 157 and converts the data into a format that can be used by the fuzzy logic controller 153. The fuzzy logic controller 1 5 3 then performs calculations to determine the fuzzy state of the particular data (f u z z y s i t u a t i ο η). Based on the above description, when the information highway begins to develop with a limited transmission rate, a method for improving the image is needed. Recently, there has been a target area (R 0 I) method that can improve image quality. However, the current R 0 I solution still has its performance obstacles. Therefore, there is a great need for a method or rule for obtaining high-quality video images. SUMMARY OF THE INVENTION In view of this, one object of the present invention is to provide a method and device that can be used to improve image quality requirements in, for example, video phone and video conference applications. In order to achieve the above and other objectives of the present invention, the present invention provides a new method and device based on the target area (R0 I) and fuzzy logic control, and will be described in detail with embodiments herein. First, the method separates a plurality of target regions of an image from a plurality of non-target regions. Next, the input from the target area is sent to a fuzzy logic control, where the fuzzy logic control is used to improve the objective.

12429TWF.PTD Page 8 200529650 V. Description of the invention (4) The quality of the target area and the improvement of the overall quality of the output image. In a preferred embodiment of the present invention, the input from the target area is calculated from the first control input and the second control input from the target area. The first control input and the second control input respectively include a first variance and a variance difference from a current i-th large block. The variance of the variance is determined by the first The number of mutations is obtained by subtracting the second variance of the previous (i-1) th large square, and dividing by the first variance. The i-th large block and the i-1 large block represent a sequence of large blocks within one of the target areas, and the i-1 large block is the previous large block of the i-th large block. In another preferred embodiment of the present invention, the fuzzy logic control includes a rule for converting control inputs into fuzzy predicates. In another preferred embodiment of the present invention, the fuzzy logic control includes a control function 'for calculating a linguistic membership function for determining the fuzzy state of the main control input. This control function uses a central area (center 0f a r e a, C 0 A) method to determine the affiliation of the language subordinate function. In another embodiment of the present invention, the fuzzy logic control includes a plurality of probe tables (100kup tables) for setting a decisional level and generating a weighted factor. Aggravate the quality of one of the target areas. In still another embodiment of the present invention, the probe tables include a plurality of

12429TWF.PTD Page 9 200529650 V. Description of the invention (5) It is used to provide a priority-like quality to one of its scaled lookup tables. The 'scaled lookup table' uses a one -Fixed and one- various slave functions formed. To sum up the above description, the present invention provides a fuzzy controlled ROI image coding. The fuzzy-controlled R0I image coding can adaptively adjust the output quality of the image: This method can easily improve the R0I quality and maintain a fixed bit device overflow (buffer 〇verfl0w), which is more familiar than conventional techniques. In addition, at a lower rate, it is easy to provide better products. ^ Codes can be used without complicated calculations, and the Shanghai Girls I will be closed together. Evening $ κυ 1 ~ Image quality. In order to make the above and other aspects of the present invention clearly understandable, the following features and advantages can be explained in more detail as follows: The attached drawings and embodiments will be described below with reference to the accompanying drawings and examples. Oral, ,, and field descriptions of the preferred implementation of the present invention are limited to these embodiments with the 0tb will be described here; may 2 "::: 眚 The present invention is not subject to the details. In the following, Both of them are familiar with the scope and components of the present invention. The corresponding reference numbers represent the same elements. First, they are composed of fuzzy control, including (1) the target area It t-shirt image coding can be controlled by two parts and (2) fuzzy control. Please refer to Figure 12429TWF.1 Page 10 200529650 V. Description of Invention (6) shows that a target area includes a segmentation unit 302. A fuzzy logic controller 3 2 0 includes a computational differential mutation unit ( calculate differential variance 303, a quantizer 304, fuzzy subsets 305, a fuzzy controller 306, a fuzzy variance operator 3 0 7, a weighted defuzzifier ( weighted deiuzzifier) 308, and a fuzzy lookup table 3 0 9. In addition, the entire encoding system also includes a Η · 2 6 3 + video encoder (video encoder) and a virtual buffer (refer to Figure 2), the fuzzy logic controller 3 2 0 improves the quality of the target area based on the difference between a variation number Ji332 and a variation number ^ (7 ^ 34). After a frame 3 0 1, the cutting unit 3 0 2 for appearance detection and motion detection will be used to cut the frame 3 0 1 into a target area (R0 1) 330 and a non-target area 331. The large block in the non-target area 331 will be directly sent to a QP selector 310 ° ROI 330 with a bit rate control without adjusting any parameters. The variation number difference Δσί334 in the i-th large block of 310 330 Calculated from aJ32 and (7/333, where σi332 and σi'333 are the variation numbers of the current and previous i-th large block respectively. The difference between the variation number Δσ i334 and the current large block variation number 332 is Use two inputs of the fuzzy logic method, and (7135 is a fuzzy output that will be used as the input weighting factor. Figures 3 and 4 show graphs representing 0 ^ 332 and △ σί334 respectively. Please refer to Figures 3 and As shown in Figure 4, the language group (linguistic

12429TWF.PTD Page 11 200529650 V. Description of the invention (7) The symbols of sets), LN351 and 401, LN352 and 402, LN353 and 403, LN354 and 404, and LN355 and 405, respectively, are "Large Positive" " , 丨 丨 Small Positive (Smal 1 Positive), 丨, Zero (Zero) 11, Small Negative (Small Negative), and "Large Negative" ". Except all σι332 are positive values, and in statistics The variation number σί334 of each large block in most of the above is outside the center of ZE3 03. The symbol in Figure 3 is exactly the same as that in Figure 4. Figure 4 shows that Δ ^ = (CTi- ( 7 /) / (Ji-defined sub-set of the number of variants. Please refer to Figure 4. As shown in Figure 4, most of the statistics △ σ 〖3 3 4 are 3 people in [一 １０ '+ 1〇] Next, the 'quantizer 304 inputs 0 ^ Mod ^ \ ΤΔ 34 into the fuzzy sub-set 30 5 and converts its degree to i ... 351 f SN 3 5 2, ZE 3 5 3, LP 3 54, And SP 3 5 5 .3 34, module, controller 3 0 6 Next by quantification ~ 3 3 2 and △ σ method, its language subordinate function 'and use the central area (C0 A ) △ σ. Two fuzzy states for 丄 1. After the calculation is completed, each of σ i /-shown in Figure 1 corresponds to the main control input value. The decision table is based on the fuzzer in Figure 5 3 π a ^ u ν,坷 仔 隹 汜 m 体 甲. Add 棹 厍 state; two Λ fuzzy probe table 309, consider ㈧ / "two kinds of W weighting factors ω ai3 3 5 to increase the RCH 3 3 0 large block. ^ ,; The monthly lookup table 3 is stored in the memory. The quality of the weighted solution is prioritized, but in one embodiment, 'in order to make different ㈧1 330 have different excellent fuzzy tables. That is, the original output of the flute is blurred and scaled (sca 1 e) — different sets of R0I priorities are used to distinguish and distinguish each and every 0ne-flxed and one-various subordinate functions of different R0I 3 3 0

200529650 V. Description of Invention (8) Example. The weighting factor is calculated in H. 2 6 3 + image encoder 31 1 using fuzzy rules for each given large block. In the experimental results of one embodiment of the present invention, it can be verified that the embodiment of the present invention has better performance than other conventionally known rules. This experiment tests Carphone, Claire, and Foreman. To define R0I in a frame, face detection is used to automatically select R0I. Compare four different methods in the test sequence. The four different methods are: encoding the frame (WR) without R0I, encoding R0I by multiplying by a weighting factor (WA), encoding R0I with three factors (TF), and the present invention (fuzzy). All four methods are set to similar average bit rates. For I-frames and P-frames with a target bit rate of 64 kbits per second, QP is set to 5 and 3, while for target I-frames with an I-signal of 32 kbits per second. For the frame and P-frame, the QP is set to 15 and 13. In WA, the weighting factor is set to 450. In TF, the three factors are set to 450, 2, and 10 respectively. In order to compare the other two methods with similar weighting, ZE13 is set to 450, and LP and LN25 are set to 350 ~ 550 °. As shown in Figure 7 to Figure 10, compared with other methods, at a similar bit rate The embodiments of the present invention have better ROI PSNR. Because both WA and TF improve the quality of ROI with fixed parameters, when the complexity of each large block changes a lot, these two methods cannot adjust their weighting factors. In summary, the embodiment of the present invention can obtain better R 0 I quality, and even when working at a lower bit rate, the phenomenon of missing frame (ski i ρ p i n g frame) will rarely occur. The invention can be applied to any image processing work, especially for immediate use.

12429TWF.PTD Page 13 200529650 V. Description of the invention (9) Image coding. Therefore, the present invention can easily improve the quality of R01 and maintain the bit rate to avoid buffer overflow. Compared with the conventional techniques, the present invention can easily improve the quality of the daylight surface with a lower bit rate. In addition, multiple R 0 I image coding can also greatly improve the quality of each R 0 I without the need for complicated operations. Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make various modifications and retouches without departing from the spirit and scope of the present invention. Therefore, the present invention The scope of protection shall be determined by the scope of the attached patent application.

12429TWF.PTD Page 14 200529650 Brief description of the diagram Figure 1a is a block diagram of a conventional feedback control rule. Figure 1b is a block diagram of a conventional fuzzy logic control law. Fig. 2 is a block diagram of image coding of a target area performed by a fuzzy logic control rule according to an embodiment of the present invention. Fig. 3 is an example of a subset of the variation number i in the fuzzy logic control device shown in Fig. 2. Fig. 4 is an example of a subset of the variation Δi in the fuzzy logic control device shown in Fig. 2. Fig. 5 is an example of a fuzzy output lookup table in the fuzzy logic control device shown in Fig. 2. Figure 6 is an example of one-fixed and one-various slave functions. Figure 7 is a comparison table of various methods for a Carphone sequence of 100 frames of 64 kbits per second. Figure 8 is a comparison table of various methods for the C 1 a i r e sequence of 150 frames of 32 kilobits per second. Figure 9 is a comparison table of the various methods of the Foreman sequence of 150 frames of 64 kbits per second. Figure 10 is a multi-target region comparison table for a NeW s sequence of 150 frames of 64 kilobits per second. Graphical label description: 1 0 0: feedback control system 1 0 1: set point

12429TWF.PTD Page 15 200529650 Simple description of the diagram 102 controller 103 processing 104 system mathematical model 105 sensor 150 fuzzy logic control system 15 1 ri-rL δ and fixed point 152 quantizer 153 fuzzy logic controller 154 defuzzifier 155 Group of rules according to human nature 156 processing 157 sensor 30 1 frame m input 302 cutting unit 303 calculation differential difference unit 304 quantizer 305 fuzzy subset 306 fuzzy controller 307 fuzzy variation number operator 308 weighted defuzzifier 309 fuzzy exploration Table 310 Weighted QP selector 31 1 Η. 2 6 3 + image encoder 312 virtual buffer

12429TWF.PTD Page 16 200529650

12429TWF.PTD Page 17

Claims (1)

200529650 6. Scope of patent application 1. An image coding method suitable for video calls and video conferences, including: separating a plurality of target areas of an image from a plurality of non-target areas; and an input from the target areas, Send to a fuzzy logic control, where the fuzzy logic control is used to improve the quality of one of the target areas and to improve the overall quality of an output image. 2. The image coding method described in item 1 of the scope of patent application, wherein the input from the target areas is calculated from a first control input and a second control input from the target areas. 3. The image coding method as described in item 2 of the scope of patent application, wherein the first control input and the second control input respectively include a first variation number and a variation number difference from a current i-th large block, The variation number difference is obtained by subtracting a first variation number from a second variation number of a previous i_l large block, and then dividing the first variation number by the first variation number. The I large block represents a sequence of the large blocks within one of the target areas, and the i-1 large block is a previous large block of the i th large block. 4. The image coding method described in item 1 of the patent application scope, wherein the fuzzy logic control includes a rule for converting the input from the target areas into a plurality of fuzzy decisions. 5. The image coding method as described in item 1 of the scope of patent application, wherein the fuzzy logic control includes a control function to calculate a language subordinate function for determining a fuzzy state. 12429TWF.PTD Page 18 200529650 6. Patent application scope 6. The image coding method described in item 5 of the patent application scope, wherein the control function includes a central area (C 0 A) method to determine the language subordinate function 〇7. The image coding method described in item 1 of the patent application scope, wherein the fuzzy logic control includes a plurality of probe tables for setting a decision level and generating a weighting factor to aggravate one of the target areas The quality of the image coding method described in item 7 of the scope of the patent application, wherein the look-up tables include a plurality of zoom look-up tables to provide a similar priority quality to one of the target areas. 9. The image coding method described in item 8 of the scope of patent application, wherein the zoom lookup tables are formed using a one-fixed and one-various subordinate function. 10. The image coding described in item 1 of the scope of patent application In the method, the fuzzy logic control further includes: · converting an input from the giant 1 marks into a plurality of fuzzy judgments;--used to determine-• each of the fuzzy judgment control functions 1 in the fuzzy state. Language subordinate function, and from this fuzzy state used to set-decision level and generate a weighting factor, a plurality of probe tables are generated to aggravate the quality of one of the target areas. 0 1 1 The image coding method according to item 10, wherein the input from the target areas is from the target areas 12429TWF.PTD Page 19 200529650 6. Scope of patent application Calculated by the first control input and a second control input. 12. The image coding method as described in item 11 of the scope of patent application, wherein the first control input and the second control input respectively include a difference between a first variation number and a variation number from an i-th block The difference between the i-th large block and the i-th large block is obtained by subtracting a second one from the first i-1 block and dividing it by the first one. The i-1 large block represents a sequence of the large blocks within one of the target areas, and the i-1 large block is a previous large block of the i-th large block. 13. The image coding method described in item 10 of the scope of patent application, wherein the control function uses a central area (CO A) method to determine the language dependent function. 14. The image coding method as described in item 10 of the scope of the patent application, wherein the lookup tables include a plurality of zoom lookup tables to provide a similar priority quality to one of the target areas. 15. The image coding method described in item 14 of the scope of the patent application, wherein the zoom lookup tables are formed using one-fixed and one-various dependent functions. 16. An image encoding device suitable for video telephone and video conference, including: an encoder having an input end and an output end, wherein the input end of the encoder is electrically coupled to an input frame; a cutting The device has an input terminal, a first output terminal and a second output terminal, wherein the input terminal of the cutting device is electrically coupled to the output terminal. 12429TWF.PTD Page 20 200529650 VI. Patent application scope; and-fuzzy logic control device with--input terminal and an output terminal, wherein the input terminal of the fuzzy logic control device is electrically coupled to the cutting device The first output terminal of the encoder, and the output terminal of the fuzzy logic control device is electrically coupled to the input terminal of the encoder. 1 7. According to the scope of patent application: the video editing described in item 16; horse device, wherein the fuzzy logic control device further includes a quantizer, which has an input terminal and an output terminal, wherein the quantizer's The input terminal is electrically coupled to the first output terminal of the cutting device, and the quantizer converts a signal 'from the first output terminal of the cutting device into a fuzzy decision;-the--controller has--the input terminal And an output terminal, wherein the input terminal of the first controller is electrically connected to the output terminal of the quantizer, the second controller converts the fuzzy decision into a fuzzy state, and the second controller, Having an input terminal and a round output terminal: wherein, the input terminal and the output terminal of the first,-'controller are electrically coupled to the output terminal of the first controller and the input terminal of the encoder, respectively; The second controller converts the logic state into an output of the fuzzy logic control device. 1 8. The image coding device described in item 17 of the scope of patent application, further includes a differential device, having an input terminal and an output terminal, wherein the input terminal and the output terminal of the differential device are electrically coupled to The first output terminal of the cutting device and the input terminal of the quantizer. 12429TWF.PTD Page 21 200529650 VI. Patent application scope 19. The image coding device described in item 18 of the patent application scope, wherein the input end of the encoder is electrically coupled to the first output end of the cutting device . 2 0. The image coding device described in item 19 of the scope of patent application, further comprising a buffer having an input end and an output end, wherein the input end and the output end of the buffer are electrically coupled respectively. To the output terminal of the encoder and the first output terminal of the cutting device. 12429TWF.PTD Page 22