Genetic Programming Survival Selection Using Intermediate Image Outputs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In the process of automatically generating image processing programs using genetic programming, effective individuals that promote learning may be eliminated, leading to an increase in time taken to generate the program due to inappropriate survival selection methods.
Innovation Solution
A program generation apparatus and method that uses genetic programming, where intermediate output images are compared with a first target image to determine if a generated program should be passed to the next generation, allowing for appropriate survival selection and reducing the time to generate an image processing program by focusing on both intermediate and final output similarities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional survival selection method comparing only final output with target image is used, then selection process is simple, but effective individuals that promote learning may be eliminated
Solution Approach 1:
The patent divides the image processing output into multiple segments: final output images and intermediate output images. By comparing both final and intermediate outputs with their respective target images, the system achieves more accurate survival selection. This segmentation allows the system to evaluate individuals at multiple stages of processing, preventing elimination of effective individuals that produce good intermediate results even if final output isn't perfect.
2Productivity
If traditional survival selection method is used, then evaluation process is fast, but learning speed decreases due to elimination of effective individuals
Solution Approach 1:
The patent performs preliminary evaluation by comparing intermediate output images with intermediate target images during the image processing workflow. This preliminary action allows early identification of effective individuals that produce correct intermediate results, enabling faster survival selection decisions without waiting for complete final output generation. This reduces the overall time to generate image processing programs by accelerating the evaluation process.
3Measurement precision
If intermediate output comparison is added to survival selection, then learning efficiency improves, but computational load increases
Solution Approach 1:
The patent applies partial action by selectively comparing only intermediate output images that are available at specific processing stages, rather than performing exhaustive comparisons of all possible outputs. This partial evaluation approach achieves sufficient precision for survival selection without the full computational burden of evaluating every possible output, thus improving learning efficiency while controlling computational energy consumption.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
To perform appropriate survival selectin when generating an image processing program by using genetic programming. A processing unit (1b) selects an image processing program (21) from among a plurality of image processing programs (21, 22, 23, and so on) each generated by combining a plurality of partial programs; generates an image processing program (21a) by changing a part of the partial programs included in the image processing program (21); performs image processing on an input image (11), using the image processing program (21a); determines whether to pass the image processing program (21a) to the next generation, based on a comparison between one or more intermediate output images (31 and 32) that are output halfway through the image processing and a first target image (12); and replaces one of the image processing programs (21, 22, 23, and so on) with the image processing program (21a) when the image processing program (21a) is determined to be passed to the next generation.