This repo is the final proj for EECS442 Computer Vision in Umich F18 titled Reconstruction and Textualization of Shredded Paper Documents by Feature Matching.
- run run_lengthwaycut on the cmd of matlab, this will provide the result on the lengthway cut approach, which will generate corresponding files on the folder.
- run run_gridcut on the cmd of matlab, this will provide the result on the grid cut approach, which will generate corresponding files on the folder.
Step0: split image in nonrandom size mode (done) and randome size mode (TBA) Step1: vertical bar reconstruction (done) Step2: laterial bar reconstruction (TBA)
- fix the bidirection problem in vertical bar reconstruction bug. Now just start from left side.
- go through the cube_cut case. Finished it in full_text case.
- Next stage is to improve the performance of reconstruction. And start contextualization (i.e. ocr).
- Add and test the matlab ocr and text saving. Finish the example. Next step is to make it a general case.
- TODO: improve the cubecut performance and prevent mismatching dimension error in programming.
//--------
- Submit your project report and the code as separate files, e.g., one pdf and one zip file. If your group created a GitHub repo, you could include the link in your report (but you still need to upload code here).
an introduction section, (documentation reconstruction significant + context) ----> dzhang
a related work section, (iteration: vertical, cube, ocr) ---> pwl
a methods section, ( method:vertical->similarity, cube->space indent, ocr->matlab build-in function ) ---> dzhang
a results section ( vertical -> niubi, [3,5] niubi, space is big-> cube niubi, space is small-> cube beng ) -----> pwl
a discussion (linear scoring, limit) ---> dzhang .
Grammer check and refine: YY & zhaoyic