Grayscale image coloring by using YCbCr and HSV color spaces

Abdul-Wahab Sami Ibrahim1 and Hind Jumaa Sartep2

1
Assist Prof. Department of computer science, College of Education, Mustansiriyah University ,Baghdad , Iraq
2
Research Scholar Department of computer science, College of Education, Mustansiriyah University ,Baghdad ,
Iraq
Abstract―Colorization is a computer aided process of adding color to a grayscale image and video her
in this paper we proposed a coloring algorithm that produce vibrant and realistic color image, generally
colorizing gray image involves greyscale image, color space and source color image. The problem of
coloring grayscale image is clearly under constrained, previous approaches have either relied on
significant user interaction her in our proposed coloring algorithm we trying to reduce the human efforts
needed in manually coloring of grayscale image, the human interaction is needed only to find a source
color image, then the job of transfer color traits from source color image to grayscale image is done by
the proposed algorithm. Here the proposed algorithm of color traits transfer to grayscale image has been
performed by using two different color spaces together in the coloring process, these color space are
YCbCr and HSV with generation of color palette for color source image and grayscale image , color
transfer depend on the comparison between the color palettes of color source image and target grayscale
image along with different pixel window size start from (2 x 2), (3 x 3) up to (10 x 10).
Key Words: Color space, grayscale image, color palette, color transfer, pixel window
I. INTRODUCTION
The colorization of grayscale images is a valuable tool for many applications such as
“colorizing” black and white films or restoring old photographs. Color can be added to grayscale images
in order to increase the visual appeal of images such as old black and white photos, classic movies or
scientific illustrations , [1], Additionally, colorization is shown to be useful in image compression,
where with the luminance information and some samples of the color (much less than the ordinary sub-
sampling in common compression schemes), the color components of the data can be faithfully
recovered[2], in medicine image modalities which only acquire grayscale image such Magnetic
Resonance Imaging (MRI) , X-ray and Computerized Tomography (CT)images can be enhanced with
color for presentations and demonstrations[3] .
The task of “colorizing” a grayscale image involves assigning three dimensional (RGB) pixel
values to an image which varies along only one dimension (luminance or intensity). Since different
colors may have the same luminance value but vary in hue or saturation, the problem of colorizing
grayscale images has no inherently “correct” solution[3] converting a color image to gray means that
dropping information about color and it is quite easy to convert color image to grayscale but its reverse is not that
easy, it look like that we can reverse the process of converting a color image to gray to get colors back, but it's not
that true. Reason for this is that there can be numerous colors which lead to one gray level but when we go reverse
of it, we cannot decide which color corresponds to this one particular gray level which we try to convert to color
[4].
Our proposed colorization method require selection of source color image whose chromatic
information is transferred to target grayscale image , unfortunately empirical evidence suggests that the
degree of similarity between source and target image has a strong influence on the quality of the result
obtain[5],after that both source and target image transfer form RGB color space to YCbCr and HSV
color space than divided both of them into pixel window size ( 2x 2) , (3 x 3) up to (10 x 10) , then start

DOI:10.21884/IJMTER.2017.4130.XD635 130
 International Journal of Modern Trends in Engineering and Research (IJMTER)
Volume 04, Issue 4, [April– 2017] ISSN (Online):2349–9745; ISSN (Print):2393-8161

compare luminance value of each pixel window form target image with all pixel window of source
image , depending on minimum Euclidean distance color component of color source image are
transferred to grayscale image pixel window found and used for colorization of respective pixel window
from input grayscale image[6].
II. COLOR SPACE
2.1 RGB color space (Standard)
The RGB model is represented by a 3-dimensional cube with red green and blue at the corners
on each axis (Figure 1). Black is at the origin. White is at the opposite end of the cube. The gray scale
follows the line from black to white. In a 24-bit color graphics system with 8 bits per color channel, red
is (255, 0, and 0). On the color cube, it is (1,0,0) [7].
The RGB model simplifies the design of computer graphics systems but it is not ideal for all
applications that red, green, and blue color components are highly correlated. This makes it difficult to
execute some image processing algorithms. Many processing techniques, such as histogram
equalization, work on the intensity component of an image only. These processes are easier
implemented using the HIS or YIQ color models [8]

Figure (1) RGB color cube[9].

2.2 YCbCr COLOR SPACE
YCbCr or sometimes written as Y′CbCr is a family of color spaces used as a part of the color
image pipeline in video and digital photography systems. Y′ is the luma component and Cb and Cr are
the blue-difference and red-difference Chroma components [5]. In the YCbCr color space, luminance
information is stored as a single component (Y), and chrominance information is stored as two color
difference components (Cb and Cr). Cb represents the difference between the blue component and a
reference value. Cr represents the difference between the red component and a reference value. [10]
To convert from RGB to YCbCr

Y= 0.299*R + 0.587*G + 0.144*B --------------- (1)

Cb=128 – 0.169*R -0.331*G + 0.500*B ---------- (2)
Cr= 128 + 0.500*R – 0.419*G – 0.081*B--------- (3)

@IJMTER-2017, All rights Reserved 131

 International Journal of Modern Trends in Engineering and Research (IJMTER)
Volume 04, Issue 4, [April– 2017] ISSN (Online):2349–9745; ISSN (Print):2393-8161

Where:
R, G, B: Red, Green and blue Channels
Y, Cb, Cr: Channels of YCbCr color space
The inverse transform equations are
R= Y + (Cb-128) + (Cr-128) ----------------------------(4)
G= Y-0.343*(Cb-128) – 0.711*(Cr-128) --------------(5)
B= Y + 1.765*(Cr-128) --------------------------------(6)
2.3 HSV COLOR SPACE
HSV is stands for (Hue, Saturation Value/Brightness). This color space is three dimensions put together gives
a volume shaped like a cone, Figure (2) illustrate the cone shape of the HSV color space, this color space has the
following characteristic [7]
 The hue (H) of a color refers to which pure color it resembles. All tints, tones and shades of red have the
same hue [11].
 The saturation (S) of a color describes how white the color is. A pure red is fully saturated, with a
saturation of 1; tints of red have saturations less than 1; and white has a saturation of 0 [11].
 The value (V) of a color, also called its lightness, describes how dark the color is. A value of 0 is black,
with increasing lightness moving away from black [11].

Figure (2) HSC color space [7]

III. PROPOSED METHOD
The following steps describe the proposed coloring methods:-
A. Select color source image that has similar features to target grayscale image.
B. Convert both color source image and target grayscale image from RGB color space to both YCbCr
and HSV color space.
C. Divide both the channels of YCbCr and HSV color space for both source and target image in to pixels
windows size (M x N).
D. Calculate the Euclidian distance from V-channel between the first pixel window form target
grayscale image with all pixel windows from V-channel of source color image separately.
E. The best matching pixel window from source color image to the first pixel window of target grayscale
image that how has the smallest Euclidian distance.

@IJMTER-2017, All rights Reserved 132

 International Journal of Modern Trends in Engineering and Research (IJMTER)
Volume 04, Issue 4, [April– 2017] ISSN (Online):2349–9745; ISSN (Print):2393-8161

F. Transfer the (Cb) and (Cr) channels from the best matching pixels window (source color image ) to
the first pixel window of target grayscale image and keep the (Y) channel of target grayscale image
unchanged.
G. Repeat steps D, E, F above for all pixels windows of target grayscale image.
H. Transfer the result image from YCbcr color space to RGB color space.
IV. QUALTY OF RESULT COLOR IMAGE
Quality of greyscale image colorization technique is subjective to the source color image
selected for coloring and also to the greyscale image to be colored. There are no objective criteria to
check the performance of colorization method. At most one may take a source greyscale of source
colorimage and try to recolor it using the colors from source color image. Then calculate the Mean
Squared Error (MSE) and peak signal to noise ratio (PSNR) between the result image and original image
could be considered serve as performance measure to see the quality of colorization method.
V. RESULT AND DISCUSSION
The grayscale coloring algorithm design using Visual Basic.net 2013 as programming language.
Our program building in personal computer has 64- bit operation system ( Microsoft Window 8.1) with
8 GB RAM and processor type Intel core ™ i7- 5500 2.4Ghz and VGA card type NVIDIA GeForce
920M.
The quality of output of colorization algorithm is subjective to the type of source color image
used to generate color palette and the target (to be colored) grayscale image. For test the performance of
the coloring proposed algorithm we used four grayscale test images as shown in Figure (3,4,5,6) are
recolored by using our coloring algorithm through (9 pixel window size) start from (2 x 2) up to (10 x
10) with calculation of mean square error (MSE) and peak signal to noise ratio (PSNR)

Figure (3):-coloring test image #1 (natural view image)

Figure (4):-coloring test image #2 (natural view with great well image

@IJMTER-2017, All rights Reserved 133

 International Journal of Modern Trends in Engineering and Research (IJMTER)
Volume 04, Issue 4, [April– 2017] ISSN (Online):2349–9745; ISSN (Print):2393-8161

Figure (5):-coloring test image #3 (animal image)

Figure (6):-coloring test image #4 (Planet image)

Below table (1) contain the values of the mean square error (MSE) and peak signal to noise ratio
(PSNR) for the above test image for pixels window size from (2 x 2) to (10 x 10).
Table (1) MSE and PSNR values for the test images

Test image #1 Test image #2 Test image #3 Test image #4

Mx N MSE PSNR MSE PSNR MSE PSNR MSE PSNR
2x2 306.83 23.261 174.933 25.702 49.764 31.161 43.121 31.783
3x3 267.203 23.862 168.034 25.876 47.378 31.374 36.887 32.462
4x4 279.973 23.659 154.052 26.254 46.525 31.543 33.967 32.82
5x5 283.152 23.61 156.803 26.177 44.635 31.633 31.745 33.114
6x6 275.333 23.732 153.945 26.257 46.02 31.501 32.763 32.976
7x7 338.32 22.837 155.523 26.212 46.9 31.419 29.777 33.391
8x8 342.939 22.778 149.753 26.377 45.12 31.587 32.053 33.072
9x9 238.414 23.606 153.083 26.253 44.936 31.604 32.992 32.946
10 x 10 266.399 23.875 160.379 26.079 43.16 31.779 33.179 32.922

below Figure (7) show the Mean Squared Error (MSE) difference between the original image
and respective re-colored image by our proposed coloring algorithm for the all test images using various
pixel window sizes, It observed that the pixel window give less MSE reflecting better coloring result
compared to the other pixel window size, as for Peak Signal to noise Ratio (PSNR) greater value
reflecting better coloring result compared with the other pixel window size see (Figure 8).

@IJMTER-2017, All rights Reserved 134

 International Journal of Modern Trends in Engineering and Research (IJMTER)
Volume 04, Issue 4, [April– 2017] ISSN (Online):2349–9745; ISSN (Print):2393-8161

MSE image#1 MSE image#2 MSE image#3 MSE image#4

400
350
300
250
MS value

200
150
100
50
0
2x2 3x3 4x4 5x5 6x6 7x7 8x8 9x9 10 x 10
Pixel Window

Figure (7):- The MSE chart for the test images

Figure (8):- PSNR chart for the all test images

VI. CONCLUSIONS
In this paper we have transfer color traits form color source image to target grayscale image by
using two different color spaces (HSV and YCbCr) with different pixels window size (2 x 2) , (3 x 3) up
to (10 x 10 ) , we conclude that the choice of source images who has similar feature to target image lead
to get better coloring result, and obtain bad coloring result in case of select unsuitable source color
image, the pixels window size give the best coloring result depend on the feature of both source color
image and target grayscale image.
REFERNCES
[1] Yogesh Rathore , Avinash Dhole , Ramnivas Giri , Umesh Agrawal , Colorization of gray scale images using Automated
Approach, IJECT Vol 1 , Issue 1 , Dec 2010.
[2] Dr. Bara'a Ali Attea, Dr. Sarab Majeed Hameed, Aminna Dahim Aboud , Content-based Grayscale Image Colorization ,
web site . http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.637.7206&rep=rep1&type=pdf
[3] H B Kekre , S D Thepade , A Athawale , Aparker , Using Assorted Color Spaces and Pixel Window Sizes for
Colorization of Grayscale images , ICWET 2010 , Mumbai , India.

@IJMTER-2017, All rights Reserved 135

 International Journal of Modern Trends in Engineering and Research (IJMTER)
Volume 04, Issue 4, [April– 2017] ISSN (Online):2349–9745; ISSN (Print):2393-8161

[4] Muhammed Imran, "Coloring gray scale image", Dec 2012, HÖGSKOLAN Dalarna.
[5] Andre Gagalowics , Wilfried Philips , Computer Analysis of Images and Patterns , 11th International , France ,
September 2005 , Springer.
[6] H B Kekre , Tanuja Sarode , Sudeep D. Thepade , Supriya Kamoji , Performance Analysis of Various Window Sizes for
Colorization of Grayscale Images using LBG and KFCG Vector Quantization Codebooks in RGB and Kekres LUV
Color Spaces , International Journal of Soft Computing and Engineering (IJSCE), ISSN: 2231-2307, Volume-1, Issue-4,
September 2011
[7] Noura abd EL-Moez AL Sebaey Semary , “Gray Image Coloring Using Texture Similarity Measure “,University of
Monofiya ,2007.
[8] Jacques Paris, web site, “Color Models for MapInfo,” April 2002 in URL, http://www.paris-pc-
gis.com/MI_Enviro/Colors/color_models.htm
[9] Web site;- https://www.researchgate.net/figure/223162168_fig5_Fig-6-The-eight-corners-of-RGB-cube
[10] Rujuta R Mahambare , " Converting Gray-Scale image to color image " International Journal of Computer, Information
Technology & Bioinformatics (IJCITB). ISSN: 2278-7593, Volume-1, Issue-4
[11] Web site http://infohost.nmt.edu/tcc/help/pubs/colortheory/web/hsv.html

@IJMTER-2017, All rights Reserved 136