2010 Second International Conference on Computer Engineering and Applications

Color Image to Grayscale Image Conversion

C. Saravanan, Ph.D., Assistant Professor

Computer Centre, National Institute of Technology,
Durgapur, West Bengal, India
[email protected]

Abstract— Conversion of a color image into a grayscale image supported by the digital device. If each Red, Green, and
inclusive of salient features is a complicated process. The Blue occupies 8 bit then the combination of RGB occupies
converted grayscale image may lose contrasts, sharpness, 24 bit and supports 16,777,216 different colors. The 24 bit
shadow, and structure of the color image. To preserve represents the color of a pixel in the color image. The
contrasts, sharpness, shadow, and structure of the color image
grayscale image has represented by luminance using 8 bits
a new algorithm has proposed. To convert the color image into
grayscale image the new algorithm performs RGB value. The luminance of a pixel value of a grayscale image
approximation, reduction, and addition of chrominance and ranges from 0 to 255. The conversion of a color image into a
luminance. The grayscale images generated using the grayscale image is converting the RGB values (24 bit) into
algorithm in the experiment confirms that the algorithm has grayscale value (8 bit).
preserved the salient features of the color image such as Various image processing techniques and software
contrasts, sharpness, shadow, and image structure. applications converts color image to grayscale image.
However, the image processing techniques or applications
Keywords: color reduction, color to gray transformation. are unable to handle the disparity in the chromaticity and the
luminance. In the literature, several linear and non-linear
I. INTRODUCTION techniques had discussed for converting color image to
Image processing is a vital research area and the grayscale image. The recent techniques handle these
utilization of images increases in various applications. On disparities much better than the earlier techniques.
different research areas, scientists are working on such as Nevertheless, the techniques involve several computational
image compression, image restoration, image segmentation procedures such as conversion of RGB space to XYZ space
etc. to enhance the existing image processing techniques and then approximations then mapping or other related
invent new method of solving image processing problems. techniques. Grayscale mappings of color images that are
The latest image processing applications such as medical constructed by approximating spectral uniformity are often
image processing, satellite image processing, and molecular inadequate [1]. The recent technique used to convert from
image processing uses various image processing techniques. color image to gray image highly consumes computational
Conversion of color image to grayscale image is one of the time and memory. Thus, a new algorithm proposed to
image processing applications used in different fields convert color image to grayscale image in a minimum
effectively. In publication organizations’ printing, a color amount of time.
image is expensive compared to a grayscale image. Thus,
II. RELATED WORK
color images have converted to grayscale image to reduce
the printing cost for low priced edition books. Similarly, There are several issues related to conversion of color
color deficient viewer requires good quality of grayscale image to grayscale image and different solutions to address
image to perceive the information, as the normal people these issues have addressed in the literature. The software
perceive the color picture. Likewise, various image such as Adobe Photoshop devised custom non-linear
processing applications require conversion of color image to projections and required users to set image dependent
grayscale image for different purpose. parameters by trial and error [2]. The following writings
Conversion of a color image to a grayscale image discuss recent six prime research works focusing on the
requires more knowledge about the color image. A pixel conversion of color image to grayscale image.
color in an image is a combination of three colors Red, A technique proposed has utilized the L*a*b luminance
Green, and Blue (RGB). The RGB color values are chrominance representation [3]. The proposed technique
represented in three dimensions XYZ, illustrated by the introduces an additive correction term for spatial
attributes of lightness, chroma, and hue. Quality of a color chrominance variations. The first step of this algorithm
image depends on the color represented by the number of computes high pass filtered versions of all three channels,
bits the digital device could support. The basic color image and the high-pass content from the two chrominance
represented by 8 bit, the high color image represented using channels combined into a single signal that represents high
16 bits, the true color image represented by 24 bit, and the frequency chrominance information. Another alternative,
deep color image is represented by 32 bit. The number of used in the implementation, is the slightly computationally
bits decides the maximum number of different colors simpler 1-norm metric. The main feature of this technique is

978-0-7695-3982-9/10 $26.00 © 2010 IEEE 196

DOI 10.1109/ICCEA.2010.192
the same color in the input image can map to different inconsistency of gradient field corrected. Finally, 2D
grayscale values depends on the spatial surround. integration applied to get the grayscale image. From the
Another novel technique proposed to handle fluorescent demonstration noted that the isoluminant colors and bluish
colors effectively [4]. Source color image converted to colors transformed to grayscale more realistic. The
uniform color space, then target differences were calculated, technique preserves overall appearance of the color image.
and finally least squares optimization technique has applied. A most recent work converted the color image and video
The experiment shows that the isoluminant colors handled to grayscale [8]. The proposed technique converted the
perfectly. The cost of setting up and solving the image and video perceptually accurate. First, H-K
optimization problem is proportional to the size of the (Helmoltz-Kohlrausch) phenomenon predicted by a
image. The proposed technique is highly resource (time and chromatic lightness term that corrects perceived lightness
memory) consumable. In addition, the technique has not based on the color’s chromatic component. The color image
provided large improvements for scenes with high dynamic converted to linear RGB by inverse gamma mapping, then
luminance range like natural scenes. transformed to CIELUV color space. Its apparent chromatic
object lightness channel calculated. Lightness channel to
A technique proposed for re-coloring of images for
grayscale values mapped using reference white chromatic
color-deficient viewers without introducing visual artifacts
values. Gamma mapping applied to move from a linear space
[5]. The mapping of color to grayscale preserves contrasts to a gamma-corrected space. Local contrast increased in the
and maintains luminance consistency. The quadratic grayscale image to represent better the local contrast of
objective function has defined for contrast preservation. original image. The work carried out using CIELab and
Further, constraints added to enforce luminance consistency CIELuv color spaces. This two step approach a good
within narrow chrominance bands. The technique performed compromise between a fully automatic technique (first step)
well for certain images and as standard for other images. and user control (second step) making this approach well
Another technique proposed enhances the contrast and suited for natural images, photographs, artistic reproductions
converts color to grayscale [6]. The proposed technique as well as business graphics. The main limitation of the
used Gaussian pairing technique for image sampling, approach is the locality of the second step. It cannot restore
dimensionality reduction, and sampling color differences. chromatic contrast between non-adjacent regions.
The predominant component analysis used for analyzing
color differences. The technique has satisfied Continuous III. ALGORITHM
mapping, Global consistency, Grayscale preservation, 1. Y = (0.299 x R) + (0.587 x G) + (0.114 x B);
Luminance ordering, Saturation ordering, and Hue ordering. 2. U = (B - Y) x 0.565;
The process controlled by three parameters: the degree of 3. V = (R - Y) x 0.713;
image enhancement; the typical size of relevant image 4. UV = U + V;
features in pixels; and the proportion of image pixels 5. R1=R*0.299;
assumed to be outliers. First, the algorithm converts the 6. R2=R*0.587;
RGB values into YPQ color space. Further, to analyze the 7. R3=R*0.114;
distribution of color contrasts between image features, color 8. G1=G*0.299;
differences between pixels considered using Gaussian 9. G2=G*0.587;
pairing. Dimensionality reduction by predominant 10. G3=G*0.114;
component analysis performed to find the color axis that 11. B1=B*0.299;
best represents the chromatic contrasts lost when the 12. B2=B*0.587;
luminance channel supplies the color to grayscale mapping. 13. B3=B*0.114;
Next, has combined luminance and chrominance 14. R4=(R1+R2+R3)/3;
information. The final step used saturation to calibrate 15. G4=(G1+G2+G3)/3;
luminance while adjusting its dynamic range and 16. B4=(B1+B2+B3)/3;
compensating for image noise. The decolorize algorithm is 17. I1=(R4+G4+B4+UV)/4;
effective at enhancing contrast. The algorithm avoids the 18. END
noise, contouring, and halo artifacts. However, tuning on
parameters required individually to suit each image. The steps 1 to 3 calculate the luminance and chrominance
A recent technique demonstrated color to grayscale values of the source color image. In the step 4 sum of
conversion based on the experimental background of the chrominance value calculated. Steps 5 to 16 the RGB values
Coloroid system observations [7]. A survey of the coloroid are approximated using RGB components. Step 17
system to and from CIE XYZ system formulas completed. calculates the average of the four values R4, G4, B4 and
Observations based on the Coloroid system discussed. The UV. The I1 represents the resulted gray color image. The
seven basic Coloroid hues fixed. Relative gray-equivalent color image, 3D plot of the RGB values of the color image,
differences of the basic hue pairs calculated. Proposed two and 3D plot of the reduced RGB values of the color image
formulas based on the CIELab color space and the Coloroid have shown and compared.
color space for building the gradient field. Further, the

197
 Table 1. Comparison of source image RGB and reduced png in type. The Mat Lab software has used for testing the
RGB values. algorithm.
The results of the experiment carefully examined. The
Source Color Image difference between the results of the proposed technique and
recent techniques identified and discussed. The result
revealed that the proposed algorithm yields grayscale image
with better luminance and chrominance property for most of
the cases and as standard for other few cases. However,
there is a minimum amount of loss in the grayscale image
due to reduction the algorithm preserved contrasts,
sharpness, shadow, and structure of the color image in the
reproduced grayscale image.
The following table shows four of the different color
3D plot of the source color image RGB images used in the experiment and respective reproduced
values grayscale images using the proposed algorithm.
Table 2. Experiment Results

Color Image Grayscale Image using

luminance components

3D plot of the reduced RGB values

The 3D plot of the source image and reduced RGB values

compared. The RGB values of the source image ranging
from 0 to 255 were reduced to a range of 0 to 85. The
reduction enhances the color range and helps to calculate the In the first grayscale image the letters A, B, and C are
grayscale in a better way. In the above picture observed that sharp as in the color image whereas it is not so in other
the color ranges are in the 3D plot of the source image RGB earlier techniques.
values are enhanced many colors are highly visible in the In the second image, the shades of the different color
3D plot of the approximated and reduced RGB values reproduced as sharp as in the color image in the grayscale
without any major changes in the colors and structure. The image than the recent algorithms.
reduction process of RGB values retained the major values V. CONCLUSION
of the RGB at most of the points observed. The reductions
made at the RGB color level so that the resultant grayscale The conversion of color image to grayscale image using
image to retain the luminance and chrominance property at the proposed algorithm uses approximation of RGB values
using luminance RGB components approximated RGB
the maximum.
reduced by three, added with chrominance value and average
IV. RESULTS AND DISCUSSION of these four value results perceptually and structurally good
quality of grayscale images. First, the luminance and
The proposed algorithm tested on thirty four number of chrominance values are calculated. Further, the RGB values
different eight bit color images published in the recent of the source color image reduced. Finally, the reduced RGB
research publications. The color values of the color images values have added chrominance values. The resulted
are in the range of 0 to 255 for each Red, Green, and Blue. grayscale images confirm that the luminance and
In the images, seven are jpg in format and twenty seven are chrominance properties and structure of the color images

198
retained well in the grayscale image. The results confirm that [4] Amy A. Gooch, Sven C. Olsen, Jack Tumblin, Bruce Gooch,
the isoluminant images have handled as handled by other Color2Gray: Salience-Preserving Color Removal, ACM
recent techniques. The proposed algorithm is helpful for SIGGRAPH, 2005, pp. 634 – 639.
different applications where good quality of grayscale image [5] K. Rasche, R. Geist, J. Westall, Re-coloring Images for Gamuts
is highly required. The proposed algorithm results best of Lower Dimension Detail Preserving Color Transformation,
quality of grayscale images using RGB reduction and EUROGRAPHICS, 24(3), 2005.
chrominance addition in a short amount of time. [6] Mark Grundland, Neil A. Dodgson, Decolorize: Fast, Contrast
Enhancing, Color to Grayscale Conversion, 40(11), 2007, pp.
REFERENCES 2891-2896.
[7] Laszlo Neumann, Martin Cadik, Antal Nemcsics, An Efficient
[1] Livingstone M. Ed., Vision and Art: The Biology of Seeing. Perception-Based Adaptive Color to Gray Transformation,
Harry N. Abrams, Inc., Publishers, 2002. EUROGRAPHICS, 2007, 73-80.
[2] Volk C., Adobe Photoshop Tip of the Week Tutorial, [8] Kaleigh Smith, Pierre-Edouard Landes, Jöelle Thollot, Karol
http://www.carlvolk.com/photoshop21.htm, 2000. Myszkowski, Apparent Greyscale: A Simple and Fast Conversion
[3] Raja Bala, Reiner Eschbach, Spatial Color-to-Grayscale to Perceptually Accurate Images and Video, 27(2), 2008.
Transform Preserving Chrominance Edge Information, Color
Imaging Conference, 2004, pp.82-86.

199