The Application Of Color Image Segmentation And Multi-Object Extraction In The Computer System Of Puzzles

Image segmentation is the principal problem of image analysis and pattern recognition,The quality of image segmentation will affect the quality of the final analysis and the result of the pattern recognition directly.So,the study of image segmentation methods has both the academic and application value.This study will pay attention on these aspects as follows:First,segment the puzzle image.Segmenting images using the fuzzy attributes can remain more image information for the latter part of image processing,so we first use fuzzy attributes for image segmentation.Experiments and analysis shows that this segmentation method is fast and its segmentation results are good.Its biggest advantage is that the segmentation results will not be affected by the number of the background color.Taking the special feature of the puzzle image into account,as the background color of the puzzle image is simple,we can complete the image segmentation by region growing based on the background.Aiming at the deficiency of the existing 4-neighbor region growing method,we propose an improved region growing method that is single direction and 2-neighbor region growing method which is based on the background of the image.This method processes only one pixel around the seed pixel,instead of processing four pixels around the seed pixel.This will greatly improve the speed of region growing.The experiment and the analysis show that on the basis of keeping basically the same segmentation result with the traditional region growing methods,the segmentation speed of the improved region growing method is faster than the speed of the traditional region growing methods.For our puzzle images,analyze and compare the segmentation method based on the fuzzy attributes and the region growing method based on the background,we can find that single direction and 2-neighbor region growing method which is based on the background of the image is a relatively simple,fast,and accurate segmentation.Second,we label the multiple target regions on the basis of the image segmentation in order to extract each interested region.The region labeling method adopted needs to judge the labels and the pixel values of the left and the above pixel of the current pixel,so that we can choose a suitable label for the current pixel.It only needs to compare 2 or 3 times,and it doesn't need extra data struct to save the label equivalent table.The speed of the labeling method improves a lot.Last,we save each labeled target region separately.In order to facilitate the next realization of the puzzle,we separately save each target region in the centre of an independent square.