Enhancing Extreme Environment Images Based On Retinal Mechanism

With the increasing applications of computer vision technology in our social life,e.g.traffic monitoring,target tracking and pattern recognition,it is important that the obtained images have high-quality as the cornerstone of the advance uses.Although there are a great number of excellent works about the image enhancement in terms of removing fog or haze,they still cannot meet the need dealing with the harsh and changeable conditions like the inhomogeneous illumination environment.It is worth noting that the frequency of the dust weather is increasing,however,there are not remarkable results of the enhancement about haze mixed with dust images.As for the research of underwater images,the demand for good algorithms will rise because of the exploring,protecting and exploiting the marine resources.Meanwhile,the medical images processing have always been the hot field in the academia and industry.Bionic intelligent algorithm based on the visual mechanism is not only effectively to address the increasing challenges,but also provides the novel techniques and ideas for the practical image engineering and computer vision.According to the retinal mechanisms,this thesis will propose three computational models for enhancing the images from the extreme environment.(1)Non-classical receptive field is a unique and key information processing mechanism in the visual system and its more peripheral region can provide the activated visual cell with obvious compensating information,which is different with the classical receptive field.From the perspective of the defogging and non-classical receptive field,this thesis will propose one simple model to enhance images.(2)The retina is a hierarchical structure,with the special feature that it is composed of both the vertical and lateral pathway.In the vertical pathway,the bipolar cells and ganglion cells have the center-periphery receptive field,which can help the retina to effectively and efficiently remove the redundant information of the visual signals.Meanwhile,the lateral pathway includes the horizontal cells and amacrine cells,and they adjust the bipolar cells and ganglion cells respectively,thereby helping to maximize the redundancy.Another novel model is based on these,which can effectively enhance various extreme images such as the fog,haze,dust,and underwater images.(3)Based on the on-pathway of the retina enhancing the edges of the pictures and the off-pathway enhancing the other information except the edges,the two channels' images processing model will be built based on the on-and off-pathways,effectively addressing the fog image and medical images about the X-ray joints.