CN110838151A

CN110838151A - Image compression processing method and device based on polynomial fitting, computer system, server and readable storage medium

Info

Publication number: CN110838151A
Application number: CN201911103016.4A
Authority: CN
Inventors: 杨帆; 周春城; 曹赛男
Original assignee: Nanjing Zhenshi Intelligent Technology Co Ltd
Current assignee: Xiaoshi Technology Jiangsu Co ltd
Priority date: 2019-11-12
Filing date: 2019-11-12
Publication date: 2020-02-25
Anticipated expiration: 2039-11-12
Also published as: CN110838151B

Abstract

The invention provides a picture compression processing method, a picture compression processing device, a computer system, a server and a readable storage medium based on polynomial fitting, wherein the picture compression processing comprises the following steps: 1) generating an average compression rate table Te; 2) fitting a polynomial nonlinear function; 3) picture coding and compression; 4) determining the closest Rx; 5) zooming the picture to be processed according to Rx; 6) and processing and outputting the zoomed picture. The invention adopts a representative Picture Group (Picture Group) under N service scenes selected for preprocessing in a preselected application scene, generates pictures corresponding to the wide-high scaling proportion, then fits a curve by solving the mean value to finally obtain a nonlinear curve which can reflect the file compression proportion and the Picture compression wide-high scaling proportion, converts the successive recursive compression problem of Rs fixation into the problem that the approximate Rs value can be calculated through a curve formula, and obviously reduces the compression times and the compression time of the pictures.

Description

Image compression processing method and device based on polynomial fitting, computer system, server and readable storage medium

Technical Field

The invention relates to the technical field of picture processing, in particular to a picture compression processing method and system based on polynomial fitting.

Background

The lossy picture compression is, for example, the picture compression in JPEG format, usually utilizes discrete cosine transform, and through the transformation from time domain to frequency domain, then filters out the high-frequency signals that are not easily perceived by human eyes, and then reduces the size of the picture file by a large margin in combination with the principle of compressing after re-encoding, and then adjusts the quality factor, that is, filters out the range of the high-frequency signals, so as to further increase the picture compression ratio, but the adoption of such a compression mode can cause more loss of picture detail information.

In the prior art, for example, the picture processing method disclosed in the 201110418364.8 chinese patent application performs color space conversion and down sampling on original image data to obtain original data based on Y, U, V color space; dividing Y, U, V each component in the raw data into separate sub-regions; performing space correlation analysis on each divided sub-region based on a space statistical model of a non-stationary process; dividing adjacent and similar sub-regions into the same class according to the correlation analysis result to perform discrete cosine transform to obtain frequency domain data; and compressing the frequency domain data. Based on a spatial statistical model of a non-stationary process and spatial correlation analysis, the distortion of a compressed picture can be relatively reduced, and the good quality of an original image can be ensured as much as possible. However, the frequency domain change adopted in the method is estimated through a statistical model, and the discrete cosine calculation times are increased, so that the change of image pixels is smoothed, and the distortion is reduced. This procedure adds an algorithm burden virtually, and introduces a sub-region partition and similarity calculation, since the parameter configuration may also cause more distortion.

In some current application scenarios, for example, a face detection scenario of the AI model, there is a high requirement for picture quality, but the requirement for resolution is not high, and generally, a face pixel is larger than 64 × 64. Therefore, for a scene with constant picture quality and a desire to reduce the picture size to reduce the processing pressure of the system, it needs to consider the implementation by reducing the picture resolution (width times height) and compressing, however, the conventional picture RAW picture compression size ratio has a non-linear relationship with the compressed picture file size, the common practice for the method of picture compression based on the fixed upper limit of size by reducing the resolution is to implement the compression by recursion, and the prior art shown in fig. 1 represents the conventional picture compression implementation flow based on the resolution recursion, which includes: setting an upper limit S1 of the picture size and a ratio Rs (Rs >0 and Rs <1) of the width and the height; decoding a current picture into a RAW picture as a reference; compressing the picture by a constant quality factor picture compression algorithm to obtain a compressed picture, wherein the file size of the compressed picture is S2; if S2> S1, performing wide-high zooming according to the zooming ratio Rs, and repeating Step 3; if S2< S1, the flow ends, resulting in a compressed picture of picture size S2.

The above solution can finally obtain a picture compression output with a file size smaller than the setting S1 by setting a suitable Rs value. However, since this flow is greatly affected by the difference between the size of the initial file S2 and the set size S1, it is easy to cause the number of recursive compressions to be uncontrolled. Meanwhile, an excessive Rs value can also cause the compression times to increase the loss operation performance and increase the compression time; an excessively small Rs value may result in a final value of S2 being much smaller than S1 due to an excessively large span granularity, which may result in an excessive reduction in picture resolution to meet the desired quality requirement, although the size limitation is achieved.

Disclosure of Invention

The invention aims to provide a picture compression processing method and system based on polynomial fitting, which solve the problem of time consumption by reducing the compression times and obtain a compressed picture meeting the requirements of size and resolution.

In order to achieve the purpose, the technical scheme adopted by the invention is as follows:

the picture compression processing method based on polynomial fitting comprises the following steps:

step 1, under a plurality of application scenes, selecting N pictures for preprocessing, generating a scaling reference table related to the stepping conditions and the width-height scaling ratio according to set stepping conditions X, and then calculating to obtain an average compression rate table Te, wherein N is a positive integer greater than or equal to 10;

step 2, on the basis of the average compression rate table Te, taking the compression rate in the compression rate table as x, taking the width-height scaling ratio as y, and performing curve fitting to obtain a quadratic polynomial:

y＝ax^2+bx+c

step3, fitting a polynomial nonlinear function by using a least square method to finally obtain a polynomial function F (x);

step 4, decoding the pre-compressed picture to obtain a RAW picture, and keeping the width and height of the picture unchanged;

step3, compressing the decoded picture through a compression function to obtain the size of the encoded file S2;

step 4, obtaining a compression rate Er of the current compressed RAW picture relative to S1, where Er is S1/S2, and S1 is a preset picture size;

step 5, taking Er as input, and calculating by using a wide high compression ratio calculation function F (x) to obtain a wide high compression ratio Rx;

step 6, using Rx to zoom the RAW picture, and compressing the RAW picture by using a picture compression function to obtain a compressed picture and a picture size S3;

and 7, judging the relation between the compressed picture size S3 and the preset picture size S1, in response to the fact that the compressed picture size S3 is larger than the preset picture size S1, taking the current compressed picture as a pre-compressed picture, repeating the steps 4-7 until the compressed picture size S3 is smaller than or equal to the preset picture size S1, and outputting the current compressed picture as output.

Further, in step 1, step conditions X are set for N pictures in a preset application scene as typical pictures, the change in the size of each picture compressed after the aspect ratio is scaled down is determined, and N compression rate tables Ts are obtained according to the ratio of the size of the compressed picture to the size of the reference picture corresponding to the step conditions X;

and obtaining a final average compression rate table Te by a term-by-term averaging method, wherein the average compression rate table Te comprises numerical values of X compression rates.

Further, in step 1, for the set step condition X, scaling each picture by an equal difference value with X as an equal division ratio based on 1, and determining a change in picture size obtained by compressing the picture according to the scaled aspect ratio.

Further, in the step 1, the step condition X is set to an integer multiple of 10.

Further, in the step 1, the step condition X is set to 100.

Further, in step 1, the polynomial function f (x) is expressed as:

F(x)＝-42.045x²+134.09x+6.7896

wherein x >0 and x < 1.

Further, the decoding in the step uses the imedecode function of opencv, and the picture compression adopts the imedecode function of opencv.

According to the second aspect of the present invention, there is also provided a picture compression processing apparatus based on polynomial fitting, including:

a module for selecting N pictures for preprocessing in a plurality of application scenes, generating a scaling reference table of which the stepping conditions are related to the width-height scaling ratio according to the set stepping condition X, and then calculating to obtain an average compression rate table Te;

a module for performing curve fitting to a quadratic polynomial by using the compression ratio in the compression rate table as x and the aspect ratio as y on the basis of the average compression rate table Te, wherein y ═ ax ^2+ bx + c;

a module for fitting a polynomial nonlinear function by using a least square method to finally obtain a polynomial function F (x);

a module for decoding the pre-compressed picture to obtain a RAW picture, wherein the width and the height of the picture are kept unchanged;

a module for compressing the decoded picture by a compression function to obtain a size of the encoded file S2;

a module for obtaining a compression ratio Er of the current compressed RAW picture to S1, wherein Er is S1/S2, and S1 is a preset picture size;

a module for calculating a wide high compression ratio Rx by taking Er as input and utilizing a wide high compression ratio calculation function F (x);

a module for scaling the RAW picture by using Rx and compressing the RAW picture by using a picture compression function to obtain a compressed picture and a picture size S3;

and a module for determining a relationship between the compressed picture size S3 and the preset picture size S1, wherein in response to the compressed picture size S3 being larger than the preset picture size S1, taking the current compressed picture as a pre-compressed picture, repeating the steps 4-7 until the compressed picture size S3 is smaller than or equal to the preset picture size S1, and outputting the current compressed picture as an output.

According to a third aspect of the present invention, there is also provided a computer system comprising:

one or more processors;

a memory storing instructions that are operable, when executed by the one or more processors, to cause the one or more processors to perform operations comprising:

step 1, under a plurality of application scenes, selecting N pictures for preprocessing, generating a scaling reference table with related stepping conditions and aspect scaling ratios according to a set stepping condition X, and then calculating to obtain an average compression rate table Te;

y＝ax^2+bx+c

According to a fourth aspect of the present invention, there is also provided a computer-readable medium storing software, the software including instructions executable by one or more computers, the instructions causing the one or more computers to perform operations by such execution, the operations comprising:

y＝ax^2+bx+c

According to a fifth aspect of the present invention, there is also provided a server having data input and output interfaces and an image processing unit, the image processing apparatus including:

Compared with the prior art, the method adopts a representative Picture Group (Picture Group) under N service scenes selected for preprocessing in a preselected application scene, generates pictures corresponding to the wide-high scaling proportion, then fits a curve by solving the mean value to finally obtain a nonlinear curve capable of reflecting the correspondence between the file compression proportion and the Picture compression wide-high scaling proportion, and converts the successive recursive compression problem of Rs fixation into the problem that the approximate Rs value can be calculated through a curve formula.

It should be understood that all combinations of the foregoing concepts and additional concepts described in greater detail below can be considered as part of the inventive subject matter of this disclosure unless such concepts are mutually inconsistent. In addition, all combinations of claimed subject matter are considered a part of the presently disclosed subject matter.

The foregoing and other aspects, embodiments and features of the present teachings can be more fully understood from the following description taken in conjunction with the accompanying drawings. Additional aspects of the present invention, such as features and/or advantages of exemplary embodiments, will be apparent from the description which follows, or may be learned by practice of specific embodiments in accordance with the teachings of the present invention.

Drawings

The drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component that is illustrated in various figures may be represented by a like numeral. For purposes of clarity, not every component may be labeled in every drawing. Embodiments of various aspects of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which:

fig. 1 is a schematic diagram of a recursive compression algorithm to implement picture compression in the prior art.

FIG. 2 is a diagram of a picture compression processing apparatus for implementing the present invention.

FIG. 3 is a flow chart of the method for compressing a picture based on polynomial fitting according to the present invention.

Fig. 4 is a diagram of the present invention for determining the average compression rate table Te.

FIG. 5 is a diagram of an exemplary hardware configuration for implementing the picture processing of the present invention;

Detailed Description

In order to better understand the technical content of the present invention, specific embodiments are described below with reference to the accompanying drawings.

In this disclosure, aspects of the present invention are described with reference to the accompanying drawings, in which a number of illustrative embodiments are shown. Embodiments of the present disclosure are not necessarily intended to include all aspects of the invention. It should be appreciated that the various concepts and embodiments described above, as well as those described in greater detail below, may be implemented in any of numerous ways, and that the concepts and embodiments disclosed herein are not limited to any embodiment. In addition, some aspects of the present disclosure may be used alone, or in any suitable combination with other aspects of the present disclosure.

The present invention is described in the following order in connection with the examples shown in fig. 2-4.

1. System configuration

2. Image processing apparatus configuration

1) Generating an average compression rate table Te;

2) fitting a polynomial nonlinear function;

3) picture coding and compression;

4) determining the closest Rx;

5) zooming the picture to be processed according to Rx;

6) processing and outputting of scaled pictures

3. Hardware configuration

[ System configuration ]

Fig. 2 exemplarily shows a system configuration for implementing the picture compression scheme of the present invention, and in conjunction with fig. 1, a system 10 for implementing a picture compression process includes a picture input interface 100, a picture processing apparatus 200, and an output interface 300.

The system 10, such as a handheld electronic device, a portable laptop computer, a desktop computer, and a server, or a camera, an online image capture device, etc., that may be used locally by a user, may also be a cloud platform, a server system, etc., that is physically separated from the user. These systems 10 may receive, via the picture input interface 100, pictures that a user desires to have a picture compression process, also referred to herein as pre-processed pictures or pre-compressed pictures.

It should be appreciated that the picture input interface 100 provides an interface for entry of picture inputs corresponding to different systems 10, such as input from a picture library of a hand-held electronic device, input of photographs taken in real time from a hand-held electronic device, a camera, etc., or input of pre-processed pictures by providing picture addresses from a network path.

The picture processing apparatus 200 receives picture data input from the picture input interface 100. The picture processing device 200 performs a picture compression processing process based on the received picture. The picture processing apparatus 200 may output the compression result through the output interface 300, for example, a laptop computer, which may output the compressed picture to a display apparatus of the device for feedback, characterization to a user, or be merged with or stored separately from the attribute information of the picture, or may upload the picture compression output to the server 400, for example, through a network interface, and/or be merged with or sent separately from the attribute information of the picture.

In some cases, the picture processing device may also forward the received original pre-processed pictures to the server 400 together.

In some embodiments, the image processing apparatus 200 may also be a server disposed at the cloud, and output the compressed image to a server for further processing through an output interface, or output the compressed image to each terminal connected to the server system through a network, such as a smart phone or a computer, through the output interface, where the terminals may be a sending terminal of the preprocessed image, or may be only a receiving terminal of the preprocessed image after compression.

[ image processing apparatus ]

Fig. 2 shows a configuration of a picture processing section in an aspect of the present disclosure, i.e., an exemplary block diagram of the aforementioned picture processing apparatus. In connection with the configuration diagram shown in fig. 2 and the process of picture processing shown in fig. 3, the picture processing apparatus includes modules that perform: 1) generating an average compression rate table Te; 2) fitting a polynomial nonlinear function; 3) picture coding and compression; 4) determining the closest Rx; 5) zooming the picture to be processed according to Rx; 6) and processing and outputting the zoomed picture.

Wherein, the processing modules of 1) to 6) can be edited by program codes of a computer, and are realized in the system shown in the above reference figure 2. The entire picture processing apparatus may be implemented in a single system, or in a combination of systems by being assigned to different apparatuses.

As shown in connection with fig. 3, an exemplary process of the picture compression process includes the following processes:

y＝ax^2+bx+c

Therefore, the contradiction between the size and the compression ratio of the picture in the picture compression processing process by the traditional recursive algorithm is changed, the invention adopts a preselected application scene to select a representative picture group (Picture group) under a plurality of N service scenes for preprocessing, generates the picture corresponding to the wide-high scaling proportion, then obtains a nonlinear curve which can reflect the correspondence between the file compression proportion and the picture compression wide-high scaling proportion by averaging and fitting the curve, and converts the successive recursive compression problem of Rs fixation into the problem that the approximate Rs value can be calculated by a curve formula.

In some exemplary schemes, in step 1, for selecting a typical picture in a preset application scene, since the purpose of selecting a picture itself is to serve as a reference, the selected typical picture should be a picture in a corresponding scene, and a picture with high resolution and high quality is selected as much as possible. For example, in the identification card recognition and zoom scene, white-background identification card photos (with personnel information pages and national emblem pages) are selected, and as the landscape photos, photos with typical landscape characteristics, such as mountains, lakes, grasslands, and the like, are selected.

In other embodiments, the method can also be obtained by a training mode, including:

randomly selecting N (N is more than or equal to 10) pictures with the resolution of more than 500 x 500 in an application scene, decoding the pictures by using a decoding function, intercepting the middle 500 x 500 pixel values, coding and storing the pictures by using a picture coding function, sorting the files according to the size, and selecting the picture with the median as a finally selected typical picture so as to go to typical pictures in the N scenes, such as 100 or more pictures.

Referring to fig. 4, for example, taking X as 100 as an example, the preselected picture is scaled from 1,0.99,0.98, …, and 0.01 according to the aspect ratio and compressed in the picture compression function, so as to obtain compressed picture sizes Sr100, Sr99, …, and Sr1, respectively.

The compression rate table at different aspect ratios is calculated using Sr100 as the reference picture size:

Ts1＝{Sr100/Sr100，Sr99/Sr100，Sr98/Sr100….Sr1/Sr100}

Ts2＝{Sr100/Sr100，Sr99/Sr100，Sr98/Sr100….Sr1/Sr100}

.

TsN＝{Sr100/Sr100，Sr99/Sr100，Sr98/Sr100….Sr1/Sr100}

and obtaining a final average compression rate table Te from T by a term-by-term averaging method.

Te＝{Rs100，Rs99，Rs98….Rs1}

Where Rsm represents the average compression ratio at the mth step, and m is 1 to 100.

Preferably, the step condition is set to an integer multiple of 10. Particularly preferably, the stepping condition X is preferably greater than or equal to 100 to facilitate obtaining a smooth-transition picture width-height scaling.

In the following embodiment, for convenience of the operation period, the stepping condition X in step 1 is set to 100.

Thus, in step 1, a polynomial nonlinear function is fitted to the sample data by the least square method, and the polynomial function f (x) is expressed as:

F(x)＝-42.045x²+134.09x+6.7896

wherein x >0 and x < 1.

Optionally, the decoding suitable for the present invention uses the imecodec function of opencv, and the picture compression uses the imecodec function of opencv.

Table 1 below is the result of picture compression by the conventional recursive algorithm of the prior art, in which the first-order aspect scaling is 75% and the upper reference limit is 30720.

TABLE 1

Table 2 shows the results of the picture compression performed by the method of the present invention

TABLE 2

The construction curve adopted by the invention is formed by fitting a sample curve under multiple scenes. As can be seen from table 2, compared with the conventional method, in two scenarios, the performance of the present invention is improved by about 5 times, and the compression deviation can be reduced by 14.69% from 22.13% to > 7.44% in comparison with the common scheme.

Therefore, through the above-described process and test results, it can be seen that the efficiency of compressing pictures based on wide-height scaling can be improved by one order of magnitude in the processing process, for a human face picture compression scene, the average compression time in the same hardware environment is reduced from 200ms to less than 30ms, the improvement effect is very obvious, and meanwhile, the picture processing scheme provided by the embodiment of the invention has better performance for a generalization scene.

[ hardware configuration ]

Fig. 5 is an example illustration showing a hardware configuration of a picture compression process according to an embodiment of the present disclosure. The picture processing apparatus 600 can implement a picture compression process such as according to an embodiment of the present disclosure.

The picture processing apparatus 600 may include a CPU 601, a ROM603, a RAM 604, a user interaction interface 609, a communication module 613, and a display 615. These components are connected to one another by, for example, a bus, and are arranged in an integrated or separate manner in a board or in an integrated circuit.

The communication module 613 may be a wired or wireless communication module, such as a 4G or 5G wireless network communication module.

The CPU 601, ROM603, and RAM 604 implement various types of functions in software by reading and executing program instructions recorded in, for example, an external memory 611. In the embodiment of the present disclosure, control of the picture processing process may be implemented by, for example, the CPU 601, the ROM603, and the RAM 604.

The user interaction interface 609 is, for example, an input device (such as a touch panel, virtual keys, an image capturing button, or the like) that receives a user operation.

The display 615 is constituted by a device capable of visually notifying a user of information. For example, the display 615 may be a display device (such as a liquid crystal display, LCD). The display 615 outputs the result output after the compression processing is implemented by the software in the CPU 601, the ROM603, and the RAM 604 as a picture, and shows it to the user.

It should be understood that each of the above-described constituent elements may be constituted by using general-purpose components, or may be constituted by hardware specialized for the function of each constituent element. This configuration may be changed as appropriate in implementation.

Embodiments of the present disclosure may include a picture processing apparatus, a system (particularly, a computer system), a picture processing method performed by the picture processing apparatus or the computer system, and a nonvolatile storage medium having an executable program recorded thereon, as described above.

The foregoing embodiments of the invention, as well as illustrated in the accompanying drawings, may be configured as follows, depending upon the specific implementation.

[ Picture compression processing device ]

A picture compression processing apparatus based on polynomial fitting, comprising:

[ computer System ]

A computer system, comprising:

one or more processors;

y＝ax^2+bx+c

[ computer-readable Medium ]

A computer-readable medium storing software comprising instructions executable by one or more computers, the instructions by such execution causing the one or more computers to perform operations comprising:

y＝ax^2+bx+c

[ Server ]

A server having data input and output interfaces and an image processing unit, the image processing apparatus comprising:

Although the present invention has been described with reference to the preferred embodiments, it is not intended to be limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, the protection scope of the present invention should be determined by the appended claims.

Claims

1. A picture compression processing method based on polynomial fitting is characterized by comprising the following steps:

y＝ax^2+bx+c

2. The method according to claim 1, wherein in step 1, a step condition X is set for N pictures in a preset application scene as typical pictures, the change in the size of each picture obtained by compressing each picture after scaling down the aspect ratio is determined, and N compression rate tables Ts are obtained according to the ratio of the size of the compressed picture to the size of the reference picture corresponding to the step condition X;

3. The method according to claim 1 or 2, wherein in step 1, for the set step condition X, scaling each picture by an equal difference value with X as an equal division ratio based on 1, and determining a change in picture size obtained by compressing the picture after scaling down the aspect ratio.

4. The method for processing picture compression based on polynomial fitting according to claim 1, wherein in step 1, the step condition X is set to an integer multiple of 10.

5. The method for processing picture compression based on polynomial fitting according to claim 1, wherein in step 1, the stepping condition X is set to 100.

6. The method for picture compression processing based on polynomial fitting according to claim 5, wherein in the step 1, the expression of the polynomial function F (x) is:

F(x)＝-42.045x²+134.09x+6.7896

wherein x >0 and x < 1.

7. The method as claimed in claim 1, wherein the decoding in the step uses opencv imecodec function, and the picture compression uses opencv imecodec function.

8. A picture compression processing apparatus based on polynomial fitting, comprising:

9. A computer system, comprising:

one or more processors;

y＝ax^2+bx+c

10. A computer-readable medium storing software, the software including instructions executable by one or more computers, the instructions by such execution causing the one or more computers to perform operations comprising:

y＝ax^2+bx+c