virtual restoration of damaged images

19
1 VIRTUAL RESTORATION OF DAMAGED IMAGES (Segmentation Based Procedure) Armenia - 2014 RUSSIAN-ARMENIAN (SLAVONIC) UNIVERSITY RESEARCH CENTER FOR CRITICAL TECHNOLOGIES

Upload: vardan-kurkchiyan

Post on 24-Jul-2015

413 views

Category:

Science


0 download

TRANSCRIPT

1

VIRTUAL RESTORATION OF DAMAGED IMAGES

(Segmentation Based Procedure)

Armenia - 2014

RUSSIAN-ARMENIAN (SLAVONIC) UNIVERSITY

RESEARCH CENTER FOR CRITICAL TECHNOLOGIES

2

Definitions

1.Let consider , for certainty, RGB color space and colore images of 24 bit format. Let intensity area (0,255) of each R,G or B color component be split into L+1 non-overlapped intervals. The intensity of components of each pixel belongs to one of these intervals. Let (kR, kG, kB) be numbers of intervals to which belongs the corresponding component of a pixel W. We say that a pixel W belongs to a color area C with interval index (kR, kG, kB), if the pixel components belong to corresponding intervals with numbers of kR, kG, kB. 2.Number of color areas equals to (L+1)3 . A segment consists of connected pixels which belong to the same color area.

3

Interface of software Image Repair

4

Description of Software IMAGE REPAIR

The Software has many operations but we consider here only three.

Operation 1. Find a segment around a fixed pixel.

Step 1-1. Enter L as integer 1, 2 or more (but less 255?)

Step 1-2. Select a pixel within image region of interest and click left

button of mouse. The procedure will find the specified segment.

5

Please click to see the result of Operation 1 (L=1)

6

Operation 2. Full segmentation and simplification.

Step 2-1. Finding all segments at fixed L. Special procedure which

performs the Operation 1 for all pixels of the image. Coordinates,

intensities and some statistical characteristics of each found

segment is memorized by computer for further processing.

Step 2-2. Simplification. Simplification is changing the intensities

of all pixels of each segment by average mean of pixels intensities

of that segment. So segment turns to monochrome structure.

7

Full segmentation and Simplification Results of Operation 2 at L=1

PSNR=23.75 | 25.01 | 25.05 dB

8

Full segmentation and Simplification Results of Operation 2 at L=2

PSNR= 24.924 | 25.94 | 26.2 dB

9

Operation 3. Transfer the averaged color of a segment to another

segment

Step 3-1. Select a pixel of necessary color and click left button of

mouse. The Operation 1 will be performed.

Step 3-2. Select a pixel within presumptive segment and click right

button of mouse. The necessary color will be transferred (see next

slide).

10

Please click to see the result of Operation 2 (L=1)

11

Please click to see the result of Operation 3 (L=1) using external color palette

12

PROCEDURE REPAIR

1. Any damaged part of an image is supposed as a segment which belongs to a color area.Thus, hunting a damaged segment leads to find a segment with appropriate L, which can be used for decision making if it is necessary to repair the damaged part.

2. Software IMAGE REPAIR allows to find and to transfer color of desired segment of various size and configuration. The process of repairing of arbitrary damaged images is very laborious but fruitful process because it allows restoration or repairing of images which can’t be achieved by using other methods.

13

Example 1

14Original image Rpaired using Software

Example 2

15

Repairing of paintings (Abegyan, 1986)

Original image Repaired using Software

16

Please click to see the process of Repairing

17

Import

Original image Repaired using Software

18

TO BE CONTINUED

19