Proposal

Image resizing has been widely used for image and video display, transmission, and analysis, etc. Several image resizing approaches have been developed. Standard image scaling is not sufficient because it uniformly change the image contents. Image cropping is also limited since the context information that may contain abundant information is abandoned during the operation. To address the issues associated with the traditional image resizing methods, Shai Avidan and Ariel Shamir proposed a seam carving method which supports content-aware image resizing for both reduction and expansion. They calculated the seams and then carved out or inserted in one direction. The seam is defined as a connected path of low energy pixels crossing the image from top to bottom, or from left to right. An image energy function is then used to define the importance of pixels. By using these operators in both directions, they can reduce and enlarge the size of the an image. However, the energy function reported in this paper still has some limitations and the performance can be further improved. In this project, we will build on and extend the the Seam Carving method reported by Shaiu Avidan and Ariel Shamir.

1 Implementing one dimension seam carving algorithm

By calculating the 8-connected path of pixels on the adjacent rows/columns, we pick the one with the lowest energy and thus has lowest impact on image to remove.

2 Implementing two dimensions seam carving algorithm & content amplification

It is a 2D version of milestone 1, however, we need to take care of the order of selecting rows and columns. Content amplification is considered as first give the image a larger scale, then run seam carving algorithm on it

3 Enlarging

Enlarging is basically a reverse version of seam carving.

4 Object removal

With a GUI, the user is allowed to give priority to certain pixels in order to keep/remove them while running with our resizing algorithm

5 Multi-Size

The user may not know the expect size of the image, it makes more sense if we can storing an explicit representation of the time-evolution implicit process of seam removals and insertions, using DP algorithm.

6 Webside and Demo

We should create an interactive online webpage for user to test and enjoy the resizing application

Time	Work
Before Feb.14	Project Proposal. Discuss about steps and timeline for the project, and applications we want to implement. Create web page
Feb.15 - Mar.01	Write code to compute the edge-based measure for an input image. Implement the dynamic programming code to find the optimal seam to remove along one direction. Compare different energy functions for content aware resizing.
Mar.02 - Apr.02	Complete two dimension seam carving. Use it to reduce the width of an image and compare it with other algorithms like Crop, Column, Pixel and Optimal carving. Complete the image enlarging and content amplification application. Write Mid-term report.
Apr.03 - Apr.15	Implement seam carving in the object removal applications.
Apr.16 - Apr.23	Implement seam carving for multi-size images, which allows the user to retarget an image indifferent size. Will work on videos if there is still time.
Apr. 24 - May 07.	Complete project webpage.