IX CONGRESO NACIONAL DEL COLOR - RUA:...

IX CONGRESO NACIONAL

DEL COLOR ALICANTE 2010

Alicante, 29 y 30 de Junio, 1 y 2 de Julio de 2010

Universidad de Alicante

PUBLICACIONES UNIVERSIDAD DE ALICANTE w

ww.

sri.u

a.es

/con

gres

os/c

olor

10

C O M I T É E S P A Ñ O L D E C O L O RS O C I E D A D E S P A Ñ O L A D E Ó P T I C A

SEDOPTICA

Publicaciones de la Universidad de AlicanteCampus de San Vicente s/n

03690 San Vicente del [email protected]

http://publicaciones.ua.esTeléfono: 965903480

Fax: 965909445

© Varios autores, 2010© de la presente edición: Universidad de Alicante

ISBN: 978-84-9717-144-1

Diseño de portada: candelaInk

Este libro ha sido debidamente examinado y valorado por evaluadores ajenos a la Universidad de Alicante, con el fin de garantizar la calidad científica del mismo.

Reservados todos los derechos. Cualquier forma de reproducción, distribución, comunicaciónpública o transformación de esta obra sólo puede ser realizada con la autorización de sus titulares,

salvo excepción prevista por la ley. Diríjase a CEDRO (Centro Español de Derechos Reprográficos,www.cedro.org) si necesita fotocopiar o escanear algún fragmento de esta obra.

IX CNC -Libro de Actas-

El IX Congreso Nacional de Color cuenta con el apoyo de las siguientes entidades:

IX CONGRESO NACIONAL DEL COLOR. ALICANTE 2010


IX Congreso Nacional de Color

Alicante,

29 y 30 de Junio, 1 y 2 de Julio


Departamento de Óptica, Farmacología y Anatomía Facultad de Ciencias

Instituto Universitario de Física Aplicada a las Ciencias y las Tecnologías (IUFACyT)


IX CNC -Libro de Actas-

IX CONGRESO NACIONAL DEL COLOR. ALICANTE 2010 COMITÉ ORGANIZADOR Presidente Francisco M. Martínez Verdú Universidad de Alicante Vicepresidente I

Vicepresidente II Secretaria Científica

Secretaria Administrativa Secretaria Técnica

Tesorero Vocal

Vocal

Vocal

Vocal Vocal

Eduardo Gilabert Pérez Joaquín Campos Acosta Esther Perales Romero Olimpia Mas Martínez

Sabrina Dal Pont

Valentín Viqueira Pérez Elísabet Chorro Calderón Verónica Marchante Bárbara Micó Vicent

Elena Marchante

Ernesto R. Baena Murillo

Universidad Politécnica de Valencia

IFA-CSIC Universidad de Alicante Universidad de Alicante Universidad de Alicante Universidad de Alicante Universidad de Alicante Universidad de Alicante Universidad de Alicante Universidad de Alicante Universidad de Alicante

COMITÉ CIENTÍFICO Natividad Alcón Gargallo Joaquín Campos Acosta

Pascual Capilla Perea Ángela García Codoner Eduardo Gilabert Pérez

José Mª González Cuasante

Francisco José Heredia Mira

Enrique Hita Villaverde Luís Jiménez del Barco Jaldo

Julio Antonio Lillo Jover

Francisco M. Martínez Verdú

Manuel Melgosa Latorre Ángel Ignacio Negueruela

Susana Otero Belmar

Jaume Pujol Ramo Javier Romero Mora

Mª Isabel Suero López

Meritxell Vilaseca Ricart

Instituto de Óptica, Color e Imagen, AIDO Instituto de Física Aplicada CSIC

Universidad de Valencia

Universidad Politécnica de Valencia Universidad Politécnica de Valencia Universidad Complutense de Madrid

Universidad de Sevilla

Universidad de Granada Universidad de Granada Universidad Complutense de Madrid Universidad de Alicante Universidad de Granada Universidad de Zaragoza

Instituto de Óptica, Color e Imagen, AIDO

Universidad Politécnica de Cataluña Universidad de Granada

Universidad de Extremadura

Universidad Politécnica de Cataluña IX CNC -Libro de Actas-

259

COLOR ACCURACY OF TWO NON-CALIBRATED COMMERCIAL CAMERAS

Martina Grosman1, Rafael Huertas2, Luis Gómez-Robledo2, Manuel Melgosa2, Samuel

Morillas3, Ana Carrasco2 1 Dept. of Textiles, Faculty of Natural Sciences and Engineering, University of Ljubljana,

Ljubljana (Slovenia). 2 Dept. of Optics, University of Granada, Granada (Spain).

3 Instituto Universitario de Matemática Pura y Aplicada, Universidad Politécnica de Valencia, Valencia (Spain).

http://www.ugr.es/~basapplcolor/, [email protected] Abstract: Obtaining color coordinates from an image is a complicated process, in which calibration and characterization of the camera are essential. However, it is interesting to study what can be obtained if the camera is not calibrated. This work achieves an approximation to this study. Two different commercial cameras have been employed to obtain images of 8 different color samples. No calibration of the cameras has been performed, recovering CIELAB coordinates of the samples through Adobe Photoshop software. The results have been compared with the CIELAB coordinates computed from the spectral radiance of the samples measured with a spectroradiometer. The illumination was controlled putting the samples in a GretagMacbeth boothlight provided with a daylight simulator lamp. In this work the aperture of the cameras was fixed in 5.6. Nevertheless the sensitivity (ISO values) and exposure time have been systematically changed. The results show differences between the cameras, the samples and the setup of sensitivity and exposure time. The average color differences of the samples are 10.83 and 18.79 CIELAB units for the two cameras with the optimal combinations of ISO value and exposure time. Keywords: Color Measurement, Digital Camera, ISO Value, Exposure Time, CIELAB.

INTRODUCTION

Color imaging is becoming more important in the measurement of color, displacing in some applications to spectroradiometers, spectrophotometers or colorimeters. Different techniques are used to recover tristimulus values from the color data saved in an image [1, 2]. All of them require the calibration and characterization of the camera used to capture the image. In this work we study if it is possible to obtain colorimetric information from a non-calibrated commercial camera, and the accuracy of the obtained data. Undeniably, calibration of the camera is necessary, and this work does not claim to avoid it [3]. Nevertheless, this work analyzes how far the non-calibrated measurements from the correct values are.

MATERIALS AND METHODS

The samples employed to collect colorimetric data were selected from a printed sheet with 33x49 colors, taken from the magazine Test Targets 6.0, page 64, published by School of Print Media, Rochester Institute of Technology in 2006, as it is shown in Fig. 1. From it, 8 basic color samples were selected, as is shown in Fig. 2. These colors are Red (31x49), Green (32x49), Blue (33x49), Cyan (28x49), Magenta (29x49), Yellow (30x49), Black (33+1x49) and White (33+1x48). The numbers in brackets are the color coordinates in the printed sheet. Alternatively it


260

would be used a standardized color chart as the GretagMacbeth ColorChecker Color Rendition (CCCR) or the GretagMacbeth ColorChecker SG (CCSG).

Figure 1. Printed sheet with 33x49 color samples.

Figure 2. Squares with the same color as the patches employed for color measurements.

Two commercial digital photographic cameras were employed in the work, a Canon EOS 350d and a Nikon D80. The Canon camera has resolution of 8 megapixels, CMOS optical sensor with size 22.2x14.8 mm and the Nikon has resolution of 10.2 megapixels and CCD optical sensor with size 23.6x15.8 mm. On Canon camera was used a 28-80 mm objective, with maximum aperture f/3.5 at 28 mm, and f/5.6 at 80 mm. On Nikon camera was used a Nikkor 18-135 mm objective, with maximum aperture f/3.5 at 18 mm, and f/5.6 at 135 mm. For both cameras, the aperture was set on the medium value 5.6 and it was not changed throughout the images acquisition process. However, different ISO values, from 100 to 1600, and exposure times, between 1/15 and 1/250 s, were employed. The photographs were saved in JPG format with a resolution of 3456x2304 (8 megapixels). The Adobe Photoshop CS3 software was used to obtain colorimetric data, CIELAB coordinates, from the photos, selecting an area of a single pixel, approximately in the center of the chip image.

Bearing in mind that in case of non-calibrated cameras the transformation used in the Adobe Photoshop software is no completely appropriated because this transformation is really device dependent; the CIELAB coordinates obtained from the images acquired with each camera have been compared with the ones computes from spectroradiometric measurements, performed with a spectroradiometer PhotoReseach SpectraScan PR-704, using a aperture of 0.125º. As perfect diffuser, a surface of PTFE11, supplied with the spectroradiometer, were used. The CIE 1931 Standard Observer has been considered in the computation of triestimulus values. Table 1 shows the CIELAB color coordinates of the 8 used samples computed from the spectroradiometric measurements.

11 PTFE corresponds to polytetrafluoroethylene, which is an organic polymer.


261

Table 1 . CIELAB color coordinates of the samples.

Red Green Blue Cyan Magenta Yellow Black White L* 59.66 63.52 43.83 67.18 59.70 89.31 24.72 99.22 a* 41.83 -39.67 19.30 -28.63 53.39 -3.68 2.24 -0.04 b* 25.00 23.14 -20.86 -34.36 6.22 63.27 3.19 0.06

C*ab 48.73 45.92 28.42 44.73 53.75 63.37 3.90 hab 30.87 149.74 312.78 230.20 6.64 93.33 54.91

Both, digital photographs and spectroradiometric measurements, were performed in a GretagMacbeth Spectralight III cabinet provide with a light-source that simulates very well the illuminant D65 [4], putting the sheet in the center of the boothlight.

RESULTS

Obviously not all the possible combinations of ISO values and exposure time are appropriated, because some combinations results in over or supra exposition of the image. Fig. 3 illustrates this statement in the case of the cyan color.

Figure 3. Images (camera Canon EOS 350d) for color cyan with different ISO values and exposure.

The optimal exposure of the photograph can be achieved with different combinations of ISO value and exposure time, as can be seen in Fig. 3. These combinations correspond to the diagonal of Fig. 3. In these optimal exposure images, the CIELAB coordinates have been obtained through the Photoshop software. Table 2 shows the color difference in CIELAB unit between each image and the CIELAB coordinates computed from the spectroradiometric measurements. Grey background denotes the lowest values for each color and camera.

Table 2 . CIELAB color differences, ∆E*ab, for optimal expositions.

Color ISO Expo sure time Red Green Blue Cyan Magenta Yellow Black White

100 1/15 7.18 12.10 10.58 7.85 10.63 4.84 35.45 0.79 200 1/30 7.56 10.19 11.82 9.92 10.84 5.41 36.54 3.52 400 1/60 7.45 10.82 12.46 8.97 12.35 5.56 39.55 3.52 800 1/125 9.32 9.34 12.36 10.10 12.94 6.20 40.43 2.29

Canon

1600 1/250 7.80 13.40 14.54 9.53 14.48 7.97 38.48 3.18 100 1/15 17.79 24.61 9.24 12.62 17.06 19.27 52.66 15.12 200 1/30 19.39 25.05 12.33 10.19 17.22 19.68 55.22 13.29 400 1/60 17.84 23.20 9.72 15.61 19.18 23.35 43.79 15.69 800 1/125 17.99 23.20 9.31 18.26 19.78 20.55 42.16 17.20

Nikon

1600 1/250 19.17 23.30 7.40 17.44 19.34 22.26 52.27 16.52


262

Table 1 shows very different color differences for each color patch. The darker colors are usually the furthest from the spectroradiometric measure. On the other hand, white and light colors have the lower color differences. Also there are important differences between the CIELAB coordinates recover from each camera, obtaining the Canon the lowest values in almost all cases. The best combination ISO value and exposure time is different for each camera and color, but most of the cases correspond to ISO 100 and exposure time 1/15 s. The lowest color differences corresponding to the optimal combinations are marked with grey background. Considering the optimal setup of sensitivity and exposure time, the average color differences of the samples are 10.83 and 18.79 CIELAB units for the Canon and Nikon cameras respectively. Similar standard deviation is achieved for the Canon (10.47 CIELAB units) versus Nikon (10.72 CIELAB units).

CONCLUSIONS

To recover properly colorimetric data from an image, calibration of the camera is unavoidable. Fixing the aperture, the optimal setup for sensitivity and exposure time depend of the camera and the color. Considering the best results, important color differences are obtained in all cases, varying from 0.76 to 42.16 CIELAB units. The results also show a clear dependency with the camera. Applying the transformation included in the Adobe Photoshop software Canon obtain better results. This transformation, which is standard and no consider calibration, cause a bias of the results, more important for the Nikon camera.

AKNOWLEGEMETNS

This research had been supported by the Egide agency (France) via the Partenariat Hubert Curien Franco-Espagnol (Program Picasso Nº 19258SF) and by the Ministerio de Ciencia e Innovación (Spain) via the Acción Integrada España-Francia (Program HF2008-0056). Also Research Project FIS2007-64266, Ministerio de Educación y Ciencia (Spain), with ERDF (European Regional Development Fund) support.

REFERENCES

[1] F. H. Imai, R. S. Berns, "Spectral Estimation Using Trichromatic Digital Cameras", Proceedings of the 1st European Conference of Colour in Graphics, Image and Vision, 492-96 (1999).

[2] E. M. Valero, J. L. Nieves, S. M. C. Nascimento, K. Amano,K., D. H. Foster, "Recovering spectral scenes with an RGB and colored filters data from natural digital camera", Color Research and Application 32, 352-60 (2007).

[3] Healey, G. E. and R. Kondepudy, "Radiometric CCD Camera Calibration and Noise Estimation", IEEE Transactions on Pattern Analysis and Machine Intelligence 16, 267-76 (1994).

[4] R. Roa, R. Huertas, M.A. López-Álvarez, L. Gómez-Robledo, M. Melgosa, “Comparación entre iluminantes y Fuentes simuladoras”, Óptica Pura y Aplicada 41, 291-300 (2008).


IX CONGRESO NACIONAL DEL COLOR - RUA:...

Documents

Transcript of IX CONGRESO NACIONAL DEL COLOR - RUA:...