Adaptive Single-pixel Imaging Based On Importance Evaluation Of Image Blocks

The target image is compressed after being sampled in traditional imaging method,which leads to the waste of sampling and storage.Single-pixel imaging combines sampling and compression only using one bucket photon detector.Adaptive sampling method is always utilized in single-pixel imaging,which the important parts of image is evaluated by thresholds through coarse image and sampled with high resolution.Adaptive single-pixel imaging is simple and fast,but high sampling rate and low imaging quality is caused by the inaccuracy of the evaluating method.To solve these problems,an adaptive single pixel imaging method based on fine granularity of image blocks and an adaptive single pixel imaging method based on entropy of gray level co-occurrence matrix of image blocks are designed to improve the image quality under the condition of similar sampling rate.Firstly,the background and significance of the research are introduced,and the technology of single pixel imaging is also described.Then,the research status of single pixel imaging technology is summarized,and the research results of single pixel imaging are classified according to different sampling methods.Finally,the existing adaptive imaging methods based on the importance prediction of image blocks are simulated and analyzed,and the problems such as inaccurate prediction,high sampling rate and low imaging quality are regarded as the research targets.An adaptive single-pixel imaging method is designed based on fine granularity classification of image blocks.Coarse image is obtained and divided into four non-overlapping blocks with the same size.Important wavelet transform coefficients of image blocks are calculated and used to classify image blocks into three types: unimportant blocks,important blocks and partly important blocks.Unimportant blocks are ignored;For important blocks,wavelet transform coefficients are sampled with higher resolution;Every partly important block is divided into four child blocks which are divided again,and all the important child blocks are sampled until there is no partly important child block.The sampling range is reduced and the imaging quality is improved under the condition of similar sampling rate.Compared with other algorithms,the average reconstruction quality of smoothed images,texture images and natural images are respectively improved by 0.4~16dB,2~5.5dB and 0.7~15dB based on the designed method.An adaptive single-pixel imaging method is designed based on the entropy of gray level co-occurrence matrix of image blocks.Coarse image is obtained and divided into multiple non-overlapping blocks with the same size.Entropy of gray level co-occurrence matrix for each block is calculated,and the complexity of image blocks is evaluated by combining entropy of neighborhood blocks.When the complexity is higher than the threshold,the corresponding image would be sampled with higher resolution.The important image information is evaluated more accurately,thus the image reconstruction quality is improved with similar sampling rate.Compared with other algorithms,the average reconstruction quality of smoothed images,texture images and natural images are respectively improved by 0.8~16dB,1.8~5.4dB and 1.5~12.7dB based on the designed method.