Skip to content

Latest commit

 

History

History
111 lines (80 loc) · 5.6 KB

README.md

File metadata and controls

111 lines (80 loc) · 5.6 KB
Raw

CV_assignment_matlab

This is a repo for Computer Vision of EI SS2022

Assignment1

Feature detection using Harris conner detection, convolution ch1

Assignment2

Find the corresponding points ch2

Assignment3

8 point aggrithm ch3, check robuts correspondence

Assigment4

3D reconstruction to find $\lambda$, plot

Tour into Picture

Group 16:

Yinghan Huang: GUI and APP design, 3 cases of vanishing point

Jingkun Feng: 3D reconstruction, homography Matrix calculation,Visual performance Optimization

Hauyu Wei, Jiawei Zou :Foreground selection and Interpolation

Qihong Zha: Interpolation and Poster

Introduction

This application can create and visualize different perspective of a room based on a single image. It can provides funktionality for the user to choose an image, select the vanishing point as well as select foreground objects. At the end, it will create a 3D tour, where the user can walk through and observe different corners of the reconstructed room by pressing a few buttons.

Dependencies

To run the application, you need to install the following Mathworks Toolboxes:

  • Image Processing Toolbox
  • Computer Vision Toolbox
  • Symbolic Math Toolbox

Quick Retart

1 vanishing point case

  1. Open Matlab and change the current folder to the project directories.
  2. Run the file main.m in Command Window.
  3. Choose an image by scolling down the selection list next to 'Choose one Image'. You can also choose images in other directories by clicking Load Custom Image.
  4. Choose different mode by clicking tha tabs 0 Vanishing Point, 1 Vanishing Point or 2 Vanishing Points.
  5. To select rear wall and to define a vanishing point of the scene, first click on Select Vanishing Point and Background, then drag the points and sides of the rectangle shown on the right
  6. (Optional) To select foreground objects, click on Select foreground. After that, you can extract the object by clicking the left mouse button to define its vertexes. Press the right mouse button to complete the extraction. Please select objects one after one. Note that our application only supports reconstructing foreground objects which are "on the floor". If you want to cancel the selected points and re-select, press the space button. If you are using Linux system, in case you pressed space but the application does not respond, click on the window bar to activate the window, and try to press space again.
  7. To start the amazing tour into the picture, click 'Start Tour'. Note that the computing time depends on the resolution of the images and also depends on whether and how many foreground objects are selected.
  8. The reconstructed 3D scene will be display in a new window. You can move the camera, rotate it and zoom by pressing some keys (See detailed instruction shown in the window).

0 Vanished Point Case

Based on the thesis, when there is no obvious vanished point in the picture, we just keep the "floor wall" and "rear wall" from the picture according to the imaginary vanished point.

  1. Choose or load images and click on tab "0 Vanished point", so that the image will be in axis (!!)
  2. Click the "Select Vanished Point and Background 0" to select the floor and rear wall in the right image axis
  3. Click the "Select Foreground 0" to select the object in a closer distance, only choose one object per click!! and right click the finish the selecting. If you want to select more than one object, just click the "Select Foreground 0" one more time
  4. Click the start tour to begin the journey

image-20220719003035420

2 Vanished Point Case

Unlike the above 2 cases, we just discuss about how to do image rectification here. Because when we "cancel" one of the vanished point, the shape will not be in shape of rectangle. So a tour will be weird.

  1. Choose or load images and click on tab "2 Vanished point"

  2. first select 2 pairs of parallel lines to do the affine rectification, also draw the vanishing line.

    it is recommended that choose the parallel lines whose intersection far away from the picture itself, otherwise the image after affine rectification will not be in the axis

  3. In the image after affine rectification, choose two perpendicular lines that you want to keep them perpendicular after metric rectification

  4. click on image rectification to see the effect.

image-20220719011018151

GUI (early version)

In our working progress, we've tried earlier the user interaction via Matlab GUI. This version provides more or less the same function as the application version, except that it cannot handle scenes with 0 or 2 vanishing points. You can find this version in the folder archiv. Please feel free to try it.

Failed Cases

There are some interesting functions that we wnat to implement but didn't succeed. Those matlab functions are stored in the folder failed_issues. Please check it out, if you have time.

References

[1] Y. Horry, K.-I. Anjyo, and K. Arai, “Tour into the picture: using a spidery mesh interface to make animation from a single image,” in Proceedings of the 24th annual conference on Computer graphics and interactive techniques, 1997, pp. 225–232. [2] K. Andersen, The geometry of an art: the history of the mathematical theory of perspective from Alberti to Monge. Springer Science & Business Media, 2008. 5 [3] A. Criminisi, P. Perez and K. Toyama, "Object removal by exemplar-based inpainting," 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings., 2003, pp. II-II, doi: 10.1109/CVPR.2003.1211538.