/Image-BG-Replacement

Multiple algorithms for Image Background Replacement

Primary LanguageJupyter Notebook

Image-BG-Replacement

Code Files multiple algorithms for Image Background Replacement implemented for the completion of CSE 610 - Advanced Multiimedia Systems at University at Buffalo for Spring 2022 by Dr. Junsong Yuan.

Problem Description

Given below is an illustration for workflow for the solution:

BG-Replacement

Implemented 3 algorithms:

  1. Otsu's thresholding
  2. Grabcut with Otsu's Thresholding Mask initialization
  3. Grabcut with K-Means segmentation Mask initialization
  4. DeepLabv3 Segmentation as Mask initialization

Results

test_1_results

test_2_results

MPEG-4 obect based coding realization through foreground extraction

Based on the results above, I proposed a way to realize MPEG-4 obect based coding

Sender

MPEG-4-Workflow-Sender

Receiver

MPEG-4-Workflow-Receiver

Example

MPEG-4-Workflow-Example