Image Matting Using Objective Decomposition and Cascading Residual Network

In this work we focus on the comprehensive review of Image Matting using concepts of Deep Learning algorithms. To achieve the objective, we have mainly focused on Neural Network based technique of Image Matting called objective decomposition (MOD Net) and Cascading Residual Network (CRN). The merging of these algorithmic approaches is performed to achieve the same goal. MOD net is used in extracting foreground objects with soft boundaries from the background image. Whereas CRN enhances the resolution of an imaging system. Till now, we have achieved the Statistical analysis on both the working models and performed experimental procedures including testing on challenging scenarios like different resolution images, low-light images, motion blurred images etc. We also worked on modifying the program by adding the logistics for foreground and background multiplication of the generated matte image which sanctioned us to change the background image of any generated matte.