Cauwerts C., Deneyer A., Bodart M., Vignetting effect of two identical fisheye lenses VIGNETTING EFFECT OF TWO IDENTICAL FISHEYE LENSES 1 2 1 Cauwerts Coralie , Deneyer Arnaud , Bodart Magali 1 2 Architecture et Climat, Université catholique de Louvain, Louvain-la-Neuve, Belgium Belgian Building Research Institute, Department Acoustics, Energy & Climate, Belgium ABSTRACT This paper aims to determine and compare the vignetting effect (luminance radial falloff) of two identical SIGMA lenses mounted on two identical cameras (same manufacturers, models and references). For the largest aperture (f/2.8), the average luminance loss of the two devices is 66%, at the periphery of the fisheye, with a difference of loss of 0.5%. For each aperture, the difference between root mean square errors (RMSE) of vignetting curve of each device is always less than 0.7%. Vignetting effect is thus similar for the two devices. 180° field and thus variation of illuminance is minimized; the optical axis of the lens does not match with the rotation axis of the camera. This device was placed under a mirror box reproducing the CIE overcast sky [4]. Two illuminance meters were placed in this mirror box in order to record the variation of horizontal illuminance. Keywords: HDR imaging, fisheye, vignetting, calibration 1. INTRODUCTION Measuring real world luminances for the entire field of view is possible with a fisheye lens and HDR imaging techniques but several calibrations are necessary to make reliable luminance measurements [1]. Pictures taken with a fisheye lens present a luminance falloff for pixels far from the center of the picture. This phenomenon called vignetting is proper to the lens and should be corrected. This paper determines vignetting effect of two devices (lens + camera) of same references to point out if vignetting differs from one device to another one. 2. MATERIALS AND METHODS Pictures were taken with two CANON EOS 40D on which two fisheye lenses SIGMA 4.5mm F2.8 EX DC HSM were mounted. Measurements were realized for four apertures (f/2.8, f/4, f/10 and f/22). The device consists in a semi circle presenting squares each 5 degrees (see fig.1). 19 white squares (ρ=78%) alternate with 18 grey ones (ρ=26%). This semi circular device differs from the one used in [1,2,3] and was preferred for several reasons: only one set of shots is necessary by aperture to obtain luminances for the Fig. 1: Device in the mirror box (plan) The camera was placed in the centre of the circle to capture the 180° field. Series of Low Dynamic Range (LDR) pictures were taken at different exposure time and then combined to reconstitute an HDR picture with the hdrgen application in Radiance. The response curve needed for each camera was determined from pictures taken with the aperture of f/22. Table 1: Settings of the camera Parameters Mode White balance Daylight Picture quality Medium/Normal Sensibility ISO 100 Number of f-stop +/- 1EV Number of shots 11 Luminance of each square was measured with a MINOLTA LS110 luminance meter, positioned at the centre of the circle. Vignetting effect was determined as the ratio between measured luminances and luminances from the calibrated HDR picture. 1 Cauwerts C., Deneyer A., Bodart M., Vignetting effect of two identical fisheye lenses Fig. 2: Luminances measurement To verify that vignetting does not depend on the color of the target, calibration factor (CF) was calculated separately for white and grey squares, for each set-up (camera + lens + aperture). For the first device, CFwhite_mean = 1.31 and CFgrey_mean = 1.21. For the second device, CFwhite_mean = 1.35 and CFgrey_mean = 1.29. For each aperture, vignetting curves, determined from grey and white squares separately, were then compared by the difference between root mean square errors (RMSE) of each curve, calculated as √∑ ( ) , where is the value of the vignetting curve, the value of the reference and N is the number of data points. The symmetry of the vignetting effect was verified in the same way. Finally, the difference of vignetting between the two devices was evaluated by this difference of RMSE too. 3. RESULTS Difference of RMSE leads to the conclusions that vignetting does not depend on the color of the target (differences respectively < 1.5% and 2.7% for each device) and was observed to be symmetric (differences < 1% for the two devices). Vignetting effect of each device is presented in fig.3. Luminance of the extreme white squares was not taken into account due to the geometrical imprecision of the fisheye picture. Vignetting effect increases with large apertures. At the periphery of the picture and for the largest aperture (f/2.8), the average luminance falloff of the two devices is maximal and reaches 66% (difference = 0.5%). For each aperture, difference between RMSE of the curves of the two devices is always less than 0.7%. 2 Fig. 3: Vignetting curves measured points (left device#1, right part: device#2) and part: For smallest apertures (f/10 and f/22), vignetting effect disappeared and an overestimation of the luminances was observed. 4. DISCUSSION The increase of vignetting effect with large apertures was already mentioned in [2]. If, theoretically, the studied fisheye lenses generate a 180° angle of view through an equisolid angle projection, measurements on the pictures showed an inaccuracy of the borderline between the circular image and its surrounding black frame. This leads to difficulties to check the adequate position of the camera to embrace exactly and symmetrically the 180° field, pointing on the central square. Another difficulty consisted in matching the luminance meter with the center of the fisheye lens which is not precisely given by the manufacturer. The observed differences of RMSE can be partially explained by these difficulties. 5. CONCLUSION AND FUTURE WORK The comparison between the two identical devices leads to the conclusion that their vignetting effect is similar (difference of RMSE < 0.7%) and is maximal at the periphery of the picture for the largest aperture (f/2.8) reaching a luminance loss around 66%. To improve the accuracy of the curves at the lens periphery and confirm that vignetting curves could be used for other devices of same characteristics (SIGMA 4.5mm F2.8 EX DC HSM mounted on CANON EOS 40D), measurements should be reiterated with an increase of the frequency of the targets at the periphery of the fisheye. Cauwerts C., Deneyer A., Bodart M., Vignetting effect of two identical fisheye lenses REFERENCES [1] D. Fan, B. Painter, J. Mardalvevic, A data collection method for long-term field studies of visual comfort in real-world daylit office environments, Proceedings of PLEA2009, Quebec City, Canada (2009) 251-256. [2] M. Inanici, Evaluation of high dynamic range photography as a luminance data acquisition system, Lighting Research and Technology 38,2 (2006) 123-136. [3] A. Jacobs, M. Wilson, Determining lens vignetting with HDR techniques, Bulgarian National Lighting Conference, Varna, Bulgaria (2007). [4] M. Bodart, A. Deneyer, A. De Herde, P. Wouters, Design of a new single-patch sky and sun simulator, Lighting Research and Technology 38,1 (2006) 73-89. ACKNOWLEDGEMENTS Coralie Cauwerts and Magali Bodart were supported by the Belgian Research National Foundation (FNRS) and Arnaud Deneyer by the Belgian Building Research Institute (BBRI). Cauwerts Coralie FNRS research fellow 1, Place du Levant 1348 Louvain-la-Neuve (Belgium) +3210/47.91.52 +3210/47.21.50 [email protected] 3
© Copyright 2026 Paperzz