Rain Streaks Detection and Removal from Color-Image Video Using Sparse Representation
Rain streaks detection and removal from color image -video is a challenging problem. The rain streak removal is considered as image denoising task. In color image –video based rain streaks removal, where the dictionary learning process can be only applied once for the first frame in a video clip of the same scene. The dictionary learning can be also used for removal of rain streaks for the succeeding frames in the clip, which is useful to both reduce the computational complexity and maintain the temporal consistency of the video. The proposed Color image - video based rain streaks removal framework based on the sparse representation. Then the high frequency part was decomposed into rain and non-rain component by using the learning sparse representation based dictionaries. To separate a rain streaks from high frequency part using the muti set feature. The multi set feature, including HOG (Histogram Oriented Gradients), DOF (Depth of Field), and eigen color. The high frequency part and muti set features is applied to remove the most rain streaks. The DOF feature is used to help to identify the main subjects to preserve in a rain image. The rain streaks are usually neutral color, where the eigen color feature is used to analyse the key features of the rain streaks. The both DOF and Eigen color features are used to identify and separate the non rain component from the misidentified rain component of an image. Our proposed framework may be also integrated with any sparse representation–based super-resolution framework to achieve super-resolution of a low quality video and noisy image video. Rain removal is a very useful and important technique in applications such as security surveillance, audio/video editing and investigations.