We report transmissive color filters based on subwavelength dielectric gratings that can replace conventional dye-based color filters used in backside-illuminated CMOS image sensor (BSI CIS) technologies. The filters are patterned in an 80-nm-thick poly-silicon film on a 115-nm-thick SiO 2 spacer layer. They are optimized for operating at the primary RGB colors, 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 exhibit peak transmittance of 60-80%, and an almost insensitive response over a ±20 • angular range. This technology enables shrinking of the pixel sizes down to near a micrometer. KeywordsSubwavelength grating, high-index contrast, color filters, CMOS image sensor Scalability of complementary metal oxide transistor (CMOS) technology has improved the performance of CMOS image sensors (CIS) in the past decade. One of the trends in the CIS technology, driven mainly by portable devices with small form-factors, is to decrease the pixel size, which will lead to improved spatial resolution for digital imaging. 1 The scalability of CIS technology has enabled lateral pixel size to be reduced from more than 10 µm to less than 2 µm in the last decade. 2,3 Along with scaling of the pixel size, there have been considerable efforts to redesign color filters, 4-6 microlenses, 7 and infrared filters 8 to prevent the degradation of the optical performance. Dye-doped polymers have been conventionally used for RGB color filters in digital color imaging. However, when the pixel size gets smaller, the optical crosstalk among pixels becomes significant because of the small absorption coefficient of the organic dyes. Besides, the dye-doped polymers are essentially photoresists that degrade under ultraviolet illumination or high temperature environments. To address these issues, plasmonic color filters made of metallic thin films with subwavelength patterning have been studied for CIS technology. 5,9,10 The plasmonic color filters have several advantages such as flexible color tunability across visible spectrum and compatibility with CMOS processes. However, the absolute efficiency of the plasmonic color filters is relatively low (40-50% range) compared to conventional organic dye-doped filters. Here, we show that dielectric subwavelength gratings can be used to achieve highly efficient transmission color filters with close to angular insensitive properties. Furthermore, compared to dye-based filters, our dielectric-based filters possess better reliability under ultraviolet illumination and at high temperature. For the optimized filter designs, we take the advantage of the backside-illumination (BSI) CIS technologies, where the color filter layer can be placed in close proximity to the photodiodes 2 16 anti-reflection coating, 17,18 and reflective-transmission filters for displays. 19 In this letter, we demonstrate designs for polarizatio...

In a previous work, it was shown that there is a curious problem with the benchmark ColorChecker dataset for illuminant estimation. To wit, this dataset has at least 3 different sets of ground-truths. Typically, for a single algorithm a single groundtruth is used. But then different algorithms, whose performance is measured with respect to different ground-truths, are compared against each other and then ranked. This makes no sense. We show in this paper that there are also errors in how each groundtruth set was calculated. As a result, all performance rankings based on the ColorChecker dataset -and there are scores of these -are inaccurate.In this paper, we re-generate a new 'recommended' groundtruth set based on the calculation methodology described by Shi and Funt. We then review the performance evaluation of a range of illuminant estimation algorithms. Compared with the legacy ground-truths, we find that the difference in how algorithms perform can be large, with many local rankings of algorithms being reversed.Finally, we draw the readers attention to our new 'open' data repository which, we hope, will allow the ColorChecker set to be rehabilitated and once again become a useful benchmark for illuminant estimation algorithms.

No abstract

The performance of the MaxRGB illumination-estimation method for color constancy and automatic white balancing has been reported in the literature as being mediocre at best; however, MaxRGB has usually been tested on images of only 8-bits per channel. The question arises as to whether the method itself is inadequate, or rather whether it has simply been tested on data of inadequate dynamic range. To address this question, a database of sets of exposure-bracketed images was created. The image sets include exposures ranging from very underexposed to slightly overexposed. The color of the scene illumination was determined by taking an extra image of the scene containing 4 Gretag Macbeth mini Colorcheckers placed at an angle to one another. MaxRGB was then run on the images of increasing exposure. The results clearly show that its performance drops dramatically when the 14-bit exposure range of the Nikon D700 camera is exceeded, thereby resulting in clipping of high values. For those images exposed such that no clipping occurs, the median error in MaxRGB's estimate of the color of the scene illumination is found to be relatively small.

The poor performance of the MaxRGB illuminationestimation method is often used in the literature as a foil when promoting some new illumination-estimation method. However, MaxRGB has usually been tested on images of only 8-bits per channel, where clipping of high radiances is likely to have occurred. The question arises as to whether the method itself is inadequate, or rather whether it has simply been tested on data of inadequate dynamic range or with inadequate preprocessing. In particular, is MaxRGB's underlying assumption that there is a white or whiteequivalent surface present in every scene too strong? This question is explored here in two ways. The first avenue of investigation is based on a new database of 105 sets of multiple-exposure images. High-dynamic range images are constructed from these sets as well. The color of the scene illumination is determined by taking an extra image of the scene containing four Gretag Macbeth mini-Colorcheckers placed at an angle to one another. MaxRGB is found to perform surprisingly well when tested on either the multipleexposure or the high-dynamic range images. The second avenue of investigation is to add some simple preprocessing to the basic MaxRGB algorithm. By removing clipped pixels followed by median filtering, MaxRGB also performs better than previously reported when tested on test images of common color constancy test sets, specifically the Simon Fraser University 321-image indoor set. In particular, the Wilcoxon signed-rank test indicates that MaxRGB outperforms the most recent bright-pixel variant of color by correlation on the 321 set. MaxRGB is also competitive against the recent Edge-Based algorithm and significantly better than the computationally intensive Bayesian method on the Grayball set and the Colorchecker set. Overall, the results presented demonstrate that MaxRGB is far more effective than it has been reputed to be. V