CN-122023186-A - File digital image processing method and system

CN122023186ACN 122023186 ACN122023186 ACN 122023186ACN-122023186-A

Abstract

The invention relates to the technical field of image sharpening, in particular to a method and a system for processing a digitized file image, which comprise the following steps: and according to the archive page scanning data, invoking a pyramid filter bank and a direction filter bank to execute non-downsampling contourlet decomposition, generating a multi-scale direction sub-band coefficient set, dividing the multi-scale direction sub-band coefficient set into overlapped sliding processing units according to preset step length and size parameters, and generating a coefficient sliding window block set. In the invention, the sharpness and contrast of the stroke edge are enhanced, and the background area coefficient is maintained or restrained, so that the synchronous amplification of noise is avoided. And carrying out inverse transformation reconstruction on the direction coefficient set subjected to texture self-adaptive enhancement processing and the low-frequency component, fusing stroke details and background intensity in the multi-direction characteristic data stream, outputting a restored image with high definition and historical texture, and improving the visual readability and the follow-up character recognition accuracy of the file digital finished product.

Inventors

WEN XIAORUI
WANG HUI
LI GUANGRUI

Assignees

新生代智能科技发展(山东)有限公司

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (9)

1. A method for processing digitized images of a file, comprising the steps of: According to the file page scanning data, invoking a pyramid filter bank and a direction filter bank to execute non-downsampling contourlet decomposition to generate a multi-scale direction sub-band coefficient set, and dividing the multi-scale direction sub-band coefficient set into overlapped sliding processing units according to preset step length and size parameters to generate a coefficient sliding window block set; Calculating a multi-level grid coverage statistical value according to the coefficient sliding window block set, and acquiring a local texture fractal dimension distribution map according to the multi-level grid coverage statistical value; According to the fractal dimension distribution diagram of the local texture, respectively comparing and screening the dimension value with a preset paper fiber coarse interval and an ink stroke edge singular interval to generate a feature classification index coordinate, substituting the feature classification index coordinate falling into the ink stroke edge singular interval into a mapping function to calculate gain weight, and calculating data at a corresponding position of the coefficient sliding window block set by using the gain weight to generate a texture self-adaptive enhancement direction coefficient set; And according to the texture self-adaptive enhancement direction coefficient set, performing matrix dimension alignment and splicing with the separated continuous low-frequency components, calling a corresponding inverse filter bank to perform non-downsampled contourlet inverse transformation reconstruction, generating a reconstructed multi-directional characteristic data stream, and fusing stroke geometric details and background intensity information contained in the reconstructed multi-directional characteristic data stream to generate a sharpness restoration archive digital image.
2. The archive digitized image processing method of claim 1 wherein the step of obtaining the set of coefficient sliding window segments comprises: According to the file page scanning data, analyzing the pixel array, marking indexes according to pyramid scale numbers and direction numbers, calling a pyramid filter bank and a direction filter bank to perform non-downsampling profile wave decomposition, distinguishing a continuous low-frequency component and a plurality of discrete high-frequency direction components, and generating a continuous low-frequency component and a plurality of discrete high-frequency direction components; Extracting a pixel brightness numerical matrix for each discrete high-frequency direction component according to the continuous low-frequency component and the plurality of discrete high-frequency direction components, copying edge rows and columns around the matrix according to the fixed pixel number to carry out boundary filling, merging coefficients of all direction components according to the scale sequence number and the direction sequence number, and generating a multi-scale direction sub-band coefficient set; Dividing the coefficients of each scale in each direction into overlapped sliding processing units according to the step length parameters and the dimension parameters according to the multi-scale direction sub-band coefficient sets, calculating the brightness value difference of adjacent pixels in each sliding processing unit according to the row-column sequence, counting the gray level change value, recording the blocking result according to the index of the sliding processing units, and generating the coefficient sliding window blocking set.
3. The archive digitized image processing method of claim 1 wherein the step of obtaining the multi-level grid overlay statistics is: and setting a grid side length pixel number set in each sliding processing unit according to the coefficient sliding window block set, dividing the sliding processing unit area into pixel grids with uniform rows and columns according to the grid side length pixel number, extracting the maximum value and the minimum value of pixel brightness values in each grid, and calculating a multi-level grid coverage statistical value.
4. The archival digital image processing method according to claim 1, wherein the step of obtaining the local texture fractal dimension distribution map comprises: calculating a local Haosdorf dimension value according to the multi-level grid coverage statistic value; And diffusing the local Hausdorff dimension value of each sliding processing unit to pixel points in the coverage range through a Gaussian weight function taking the center of the sliding processing unit as a peak value, accumulating the local Hausdorff dimension values of all the sliding processing units covering the current pixel point for each pixel point, and generating a local texture fractal dimension distribution diagram.
5. The archive digitized image processing method of claim 1 wherein the feature classification index coordinate obtaining step is: And according to the local texture fractal dimension distribution diagram, reading dimension values pixel by pixel, loading a paper fiber coarse section and an ink stroke edge singular section, comparing whether the dimension values fall into the paper fiber coarse section or the ink stroke edge singular section according to pixel coordinates, marking pixel coordinates falling into the ink stroke edge singular section, recording category identifiers, and generating feature classification index coordinates.
6. The archive digitized image processing method of claim 1 wherein the step of obtaining the texture adaptive enhancement direction coefficient set comprises: according to the feature classification index coordinates, screening pixel coordinates marked as falling into a singular interval of the ink stroke edge, reading dimension numbers corresponding to the pixel coordinates according to a mapping function, performing input and output conversion, establishing a corresponding relation between the pixel coordinates and conversion results according to the pixel coordinates, storing indexes, and generating gain weights; And positioning the direction coefficient position of each pixel coordinate in the coefficient sliding window partition set, obtaining an original direction coefficient according to the partition number and the direction number of the pixel coordinate, scaling the original direction coefficient by bits by using the corresponding gain weight, averaging the overlapping positions, keeping the direction coefficient of the unselected pixel coordinate unchanged, and generating a texture self-adaptive enhancement direction coefficient set.
7. The archive digitized image processing method of claim 1 wherein said reconstructing multidirectional feature data stream comprises the steps of: According to the texture self-adaptive enhancement direction coefficient set, reading coefficient matrixes according to scale numbers and direction numbers, performing row-column dimension comparison with separated continuous low-frequency components, calculating filling values required by the continuous low-frequency components, completing pixel-level filling at four-side copying edge rows and columns, splicing the continuous low-frequency components and the direction coefficient matrixes in channel dimension according to the scale numbers and the direction numbers, and generating matrix dimension alignment and spliced coefficient combinations; and selecting the filter corresponding to the inverse filter group according to the scale number and the direction number according to the coefficient combination after the matrix dimension alignment and the splicing, inputting the inverse filter group by group to carry out inverse non-downsampled contourlet inverse transformation reconstruction, outputting a direction response sequence arranged according to the input sequence, and generating a reconstructed multi-directional characteristic data stream.
8. The archive digitized image processing method of claim 1 wherein said sharpness restoration archive digital image acquisition step is: and separating stroke geometric detail response and background intensity information according to the reconstructed multidirectional characteristic data stream, carrying out addition fusion according to pixel coordinates, cutting off abnormal values in a pixel value range, obtaining an inverse transformed pixel matrix, writing the inverse transformed pixel matrix into an image buffer area, and generating a sharpness restoration archive digital image.
9. A system for archival digital image processing methods according to any one of claims 1-8, comprising: the image decomposition module is used for calling the pyramid filter bank and the direction filter bank to execute non-downsampling contourlet decomposition according to the archive page scanning data, generating a multi-scale direction sub-band coefficient set, dividing the multi-scale direction sub-band coefficient set into overlapped sliding processing units according to a preset step length and a preset size parameter, and generating a coefficient sliding window block set; The characteristic extraction module is used for calculating a multi-level grid coverage statistical value according to the coefficient sliding window block set and obtaining a local texture fractal dimension distribution diagram according to the multi-level grid coverage statistical value; The ink mark enhancement module is used for comparing and screening the dimension values with a preset paper fiber coarse interval and an ink mark stroke edge singular interval respectively according to the local texture fractal dimension distribution diagram to generate feature classification index coordinates, substituting the feature classification index coordinates falling into the ink mark stroke edge singular interval into a mapping function to calculate gain weights, and calculating data at positions corresponding to the coefficient sliding window block set by using the gain weights to generate a texture self-adaptive enhancement direction coefficient set; and the image reconstruction module is used for carrying out matrix dimension alignment and splicing on the texture self-adaptive enhancement direction coefficient set and the separated continuous low-frequency components, calling a corresponding inverse filter bank to carry out non-downsampled contourlet inverse transformation reconstruction, generating a reconstructed multi-direction characteristic data stream, and fusing stroke geometric details and background intensity information contained in the reconstructed multi-direction characteristic data stream to generate a sharpness restoration archive digital image.

Description

File digital image processing method and system Technical Field The invention relates to the technical field of image sharpening, in particular to a method and a system for processing a digitized file image. Background Image sharpening is one of the techniques used in digital image processing to improve image quality, and its main purpose is to compensate for edge blurring and detail loss problems caused by optical system aberrations, motion blur, system resolution limitations, or ambient noise interference during digital acquisition (e.g., scanner scanning, camera shooting) or transmission of images. In the prior art, when processing an image sharpening task, a global unified sharpening operator or a gradient-based anti-sharpening mask algorithm is generally relied on, and edges are mainly highlighted by enhancing high-frequency parts with intense gray level jump in an image. When the method is practically applied to a history file digital scene, due to physical characteristics of aging materials, rough fibers, macula lutea or folds and the like of file paper, the background textures are also expressed as high-frequency signals after digital scanning. When the sharpening operation is executed, paper fiber noise can be amplified indiscriminately, so that obvious granular feel or artifacts appear on the processed image background, and the original historical appearance of the file is destroyed. Therefore, improvements are needed. Disclosure of Invention The invention aims to solve the defects in the prior art and provides a method and a system for processing a digitized file image. In order to achieve the above purpose, the invention adopts the following technical scheme that the archive digitized image processing method comprises the following steps: According to the file page scanning data, invoking a pyramid filter bank and a direction filter bank to execute non-downsampling contourlet decomposition to generate a multi-scale direction sub-band coefficient set, and dividing the multi-scale direction sub-band coefficient set into overlapped sliding processing units according to preset step length and size parameters to generate a coefficient sliding window block set; Calculating a multi-level grid coverage statistical value according to the coefficient sliding window block set, and acquiring a local texture fractal dimension distribution map according to the multi-level grid coverage statistical value; According to the fractal dimension distribution diagram of the local texture, respectively comparing and screening the dimension value with a preset paper fiber coarse interval and an ink stroke edge singular interval to generate a feature classification index coordinate, substituting the feature classification index coordinate falling into the ink stroke edge singular interval into a mapping function to calculate gain weight, and calculating data at a corresponding position of the coefficient sliding window block set by using the gain weight to generate a texture self-adaptive enhancement direction coefficient set; And according to the texture self-adaptive enhancement direction coefficient set, performing matrix dimension alignment and splicing with the separated continuous low-frequency components, calling a corresponding inverse filter bank to perform non-downsampled contourlet inverse transformation reconstruction, generating a reconstructed multi-directional characteristic data stream, and fusing stroke geometric details and background intensity information contained in the reconstructed multi-directional characteristic data stream to generate a sharpness restoration archive digital image. Preferably, the step of obtaining the coefficient sliding window block set includes: According to the file page scanning data, analyzing the pixel array, marking indexes according to pyramid scale numbers and direction numbers, calling a pyramid filter bank and a direction filter bank to perform non-downsampling profile wave decomposition, distinguishing a continuous low-frequency component and a plurality of discrete high-frequency direction components, and generating a continuous low-frequency component and a plurality of discrete high-frequency direction components; Extracting a pixel brightness numerical matrix for each discrete high-frequency direction component according to the continuous low-frequency component and the plurality of discrete high-frequency direction components, copying edge rows and columns around the matrix according to the fixed pixel number to carry out boundary filling, merging coefficients of all direction components according to the scale sequence number and the direction sequence number, and generating a multi-scale direction sub-band coefficient set; Dividing the coefficients of each scale in each direction into overlapped sliding processing units according to the step length parameters and the dimension parameters according to the multi-scale direction sub-band coefficient sets, calcula