PROJECT TITLE :
Dynamic Scene Deblurring by Depth Guided Model
Object movement, depth fluctuation, and camera shake are the most common causes of dynamic scene blur. For the most part, present approaches use picture segmentation or trainable deep convolutional neural networks that take into account both the motion of the objects and the shake of the camera to resolve this issue. These methods, however, are less effective when there are deep fluctuations in the data. For dynamic scene deblurring, we propose a deep neural convolutional network that uses the depth map The edges and structure of the depth map are restored using a depth refinement network after the depth map has been extracted from a blurred image. It is necessary to employ the spatial feature transform layer in order to obtain depth features from the depth map, which are then combined with the image features by scaling and shifting. With the help of the depth map, our picture deblurring network learns how to recover a clear image. For the suggested model, depth information is critical, according to extensive experimentation and research. The suggested model has also been evaluated quantitatively and qualitatively, and the results show that it outperforms current dynamic scene deblurring approaches and traditional depth-based deblurring algorithms.
Did you like this research project?
To get this research project Guidelines, Training and Code... Click Here