/** @file sanei_ir.h * * This file provides an interface to the * sanei_ir functions for utilizing the infrared plane * * Copyright (C) 2012 Michael Rickmann * * This file is part of the SANE package. * * Essentially three things have to be done: * - 1) reduce red spectral overlap from the infrared (ired) plane * - 2) find the dirt * - 3) replace the dirt * * - 1) is mainly adressed by sanei_ir_spectral_clean * - 2) by sanei_ir_filter_madmean * - 3) by sanei_ir_dilate_mean */ #ifndef SANEI_IR_H #define SANEI_IR_H #include #define SAMPLE_SIZE 40000 /**< maximal for random sampling */ #define HISTOGRAM_SHIFT 8 /**< standard histogram size */ #define HISTOGRAM_SIZE (1 << HISTOGRAM_SHIFT) #define SAFE_LOG(x) ( ((x) > 0.0) ? log ((x)) : (0.0) ) /**< define log (0) = 0 */ #define MAD_WIN2_SIZE(x) ( (((x) * 4) / 3) | 1 ) /**< MAD filter: 2nd window size */ typedef uint16_t SANE_Uint; /** * @brief Pointer to access values of different bit depths */ typedef union { uint8_t *b8; /**< <= 8 bits */ uint16_t *b16; /**< > 8, <= 16 bits */ } SANEI_IR_bufptr; /** Initialize sanei_ir. * * Call this before any other sanei_ir function. */ extern void sanei_ir_init (void); /** * @brief Create the normalized histogram of a grayscale image * * @param[in] params describes image * @param[in] img_data image pointer { grayscale } * @param[out] histogram an array of double with histogram * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * * @note * histogram has to be freed by calling routine */ extern SANE_Status sanei_ir_create_norm_histogram (const SANE_Parameters * params, const SANE_Uint *img_data, double ** histogram); /** * @brief Implements Yen's thresholding method * * @param[in] params describes image * @param[in] norm_histo points to a normalized histogram * @param[out] thresh found threshold * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * * -# Yen J.C., Chang F.J., and Chang S. (1995) "A New Criterion * for Automatic Multilevel Thresholding" IEEE Trans. on Image * Processing, 4(3): 370-378 * -# Sezgin M. and Sankur B. (2004) "Survey over Image Thresholding * Techniques and Quantitative Performance Evaluation" Journal of * Electronic Imaging, 13(1): 146-165 * -# M. Emre Celebi, 06.15.2007, fourier_0.8, * http://sourceforge.net/projects/fourier-ipal/ * -# ImageJ Multithresholder plugin, * http://rsbweb.nih.gov/ij/plugins/download/AutoThresholder.java */ extern SANE_Status sanei_ir_threshold_yen (const SANE_Parameters * params, double * norm_histo, int *thresh); /** * @brief Implements Otsu's thresholding method * * @param[in] params describes image * @param[in] norm_histo points to a normalized histogram * @param[out] thresh found threshold * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * * -# Otsu N. (1979) "A Threshold Selection Method from Gray Level Histograms" * IEEE Trans. on Systems, Man and Cybernetics, 9(1): 62-66 * -# M. Emre Celebi, 06.15.2007, fourier_0.8 * http://sourceforge.net/projects/fourier-ipal/ */ extern SANE_Status sanei_ir_threshold_otsu (const SANE_Parameters * params, double * norm_histo, int *thresh); /** * @brief Implements a Maximum Entropy thresholding method * * @param[in] params describes image * @param[in] norm_histo points to a normalized histogram * @param[out] thresh found threshold * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * * -# Kapur J.N., Sahoo P.K., and Wong A.K.C. (1985) "A New Method for * Gray-Level Picture Thresholding Using the Entropy of the Histogram" * Graphical Models and Image Processing, 29(3): 273-285 * -# M. Emre Celebi, 06.15.2007, fourier_0.8 * http://sourceforge.net/projects/fourier-ipal/ * -# ImageJ Multithresholder plugin, * http://rsbweb.nih.gov/ij/plugins/download/AutoThresholder.java */ extern SANE_Status sanei_ir_threshold_maxentropy (const SANE_Parameters * params, double * norm_histo, int *thresh); /** * @brief Generate gray scale luminance image from separate R, G, B images * * @param params points to image description * @param[in] in_img pointer to at least 3 planes of image data * @param[out] out_img newly allocated image * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * - SANE_STATUS_UNSUPPORTED - wrong input bit depth * * @note out_img has to be freed by the calling routine. * @note on input params describe a single color plane, * on output params are updated if image depth is scaled */ SANE_Status sanei_ir_RGB_luminance (SANE_Parameters * params, const SANE_Uint **in_img, SANE_Uint **out_img); /** * @brief Convert image from >8 bit depth to an 8 bit image. * * @param[in] params pimage description * @param[in] in_img points to input image data * @param[out] out_params if != NULL * receives description of new image * @param[out] out_img newly allocated 8-bit image * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * - SANE_STATUS_UNSUPPORTED - wrong input bit depth * * @note * out_img has to be freed by the calling routine, */ extern SANE_Status sanei_ir_to_8bit (SANE_Parameters * params, const SANE_Uint *in_img, SANE_Parameters * out_params, SANE_Uint **out_img); /** * @brief Allocate and initialize logarithmic lookup table * * @param[in] len length of table, usually 1 << depth * @param[out] lut_ln adress of pointer to allocated table * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * * @note natural logarithms are provided */ SANE_Status sanei_ir_ln_table (int len, double **lut_ln); /** * @brief Reduces red spectral overlap from an infrared image plane * * @param[in] params pointer to image description * @param[in] lut_ln pointer lookup table * if NULL it is dynamically handled * @param[in] red_data pointer to red image plane * @param ir_data pointer to ir image plane * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * * This routine is based on the observation that the relation beween the infrared value * ired and the red value red of an image point can be described by ired = b + a * ln (red). * First points are randomly sampled to calculate the linear regression coefficent a. * Then ired' = ired - a * ln (red) is calculated for each pixel. Finally, the ir' image * is scaled between 0 and maximal value. For the logarithms a lookup table is used. * Negative films show very little spectral overlap but positive film usually has to be * cleaned. As we do a statistical measure of the film here dark margins and lumps of * dirt have to be excluded. * * @note original ired data are replaced by the cleaned ones */ extern SANE_Status sanei_ir_spectral_clean (const SANE_Parameters * params, double *lut_ln, const SANE_Uint *red_data, SANE_Uint *ir_data); /** * @brief Optimized mean filter * * @param[in] params pointer to image description * @param[in] in_img Pointer to grey scale image data * @param[out] out_img Pointer to grey scale image data * @param[in] win_rows Height of filtering window, odd * @param[in] win_cols Width of filtering window, odd * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * - SANE_STATUS_INVAL - wrong window size * * @note At the image margins the size of the filtering window * is adapted. So there is no need to pad the image. * @note Memory for the output image has to be allocated before */ extern SANE_Status sanei_ir_filter_mean (const SANE_Parameters * params, const SANE_Uint *in_img, SANE_Uint *out_img, int win_rows, int win_cols); /** * @brief Find noise by adaptive thresholding * * @param[in] params pointer to image description * @param[in] in_img pointer to grey scale image * @param[out] out_img address of pointer to newly allocated binary image * @param[in] win_size Size of filtering window * @param[in] a_val Parameter, below is definetly clean * @param[in] b_val Parameter, above is definetly noisy * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * * This routine follows the concept of Crnojevic's MAD (median of the absolute deviations * from the median) filter. The first median filter step is replaced with a mean filter. * The dirty pixels which we wish to remove are always darker than the real signal. But * at high resolutions the scanner may generate some noise and the ired cleaning step can * reverse things. So a maximum filter will not do. * The second median is replaced by a mean filter to reduce computation time. Inspite of * these changes Crnojevic's recommendations for the choice of the parameters "a" and "b" * are still valid when scaled to the color depth. * * @reco Crnojevic recommends 10 < a_val < 30 and 50 < b_val < 100 for 8 bit color depth * * @note a_val, b_val are scaled by the routine according to bit depth * @note "0" in the mask output is regarded "dirty", 255 "clean" * * -# Crnojevic V. (2005) "Impulse Noise Filter with Adaptive Mad-Based Threshold" * Proc. of the IEEE Int. Conf. on Image Processing, 3: 337-340 */ extern SANE_Status sanei_ir_filter_madmean (const SANE_Parameters * params, const SANE_Uint *in_img, SANE_Uint ** out_img, int win_size, int a_val, int b_val); /** * @brief Add dark pixels to mask from static threshold * * @param[in] params pointer to image description * @param[in] in_img pointer to grey scale image * @param mask_img pointer to binary image (0, 255) * @param[in] threshold below which the pixel is set 0 */ void sanei_ir_add_threshold (const SANE_Parameters * params, const SANE_Uint *in_img, SANE_Uint * mask_img, int threshold); /** * @brief Calculates minimal Manhattan distances for an image mask * * @param[in] params pointer to image description * @param[in] mask_img pointer to binary image (0, 255) * @param[out] dist_map integer pointer to map of closest distances * @param[out] idx_map integer pointer to indices of closest pixels * @param[in] erode == 0: closest pixel has value 0, != 0: is 255 * * manhattan_dist takes a mask image consisting of 0 or 255 values. Given that * a 0 represents a dirty pixel and erode != 0, manhattan_dist will calculate the * shortest distance to a clean (255) pixel and record which pixel that was so * that the clean parts of the image can be dilated into the dirty ones. Thresholding * can be done on the distance. Conversely, if erode == 0 the distance of a clean * pixel to the closest dirty one is calculated which can be used to dilate the mask. * * @ref extended and C version of * http://ostermiller.org/dilate_and_erode.html */ void sanei_ir_manhattan_dist (const SANE_Parameters * params, const SANE_Uint * mask_img, unsigned int *dist_map, unsigned int *idx_map, unsigned int erode); /** * @brief Dilate or erode a mask image * * @param[in] params pointer to image description * @param mask_img pointer to binary image (0, 255) * @param dist_map integer pointer to map of closest distances * @param idx_map integer pointer to indices of closest pixels * @param[in] by number of pixels, > 0 dilate, < 0 erode * * @note by > 0 will enlarge the 0 valued area */ void sanei_ir_dilate (const SANE_Parameters * params, SANE_Uint * mask_img, unsigned int *dist_map, unsigned int *idx_map, int by); /** * @brief Suggest cropping for dark margins of positive film * * @param[in] params pointer to image description * @param[in] dist_map integer pointer to map of closest distances * @param[in] inner crop within (!=0) or outside (==0) the image's edges * @param[out] edges pointer to array holding top, bottom, left * and right edges * * The distance map as calculated by sanei_ir_manhattan_dist contains * distances to the next clean pixel. Dark margins are detected as dirt. * So the first/last rows/columns tell us how to crop. This is rather * fast if the distance map has been calculated anyhow. */ void sanei_ir_find_crop (const SANE_Parameters * params, unsigned int * dist_map, int inner, int * edges); /** * @brief Dilate clean image parts into dirty ones and smooth int inner, * * @param[in] params pointer to image description * @param in_img array of pointers to color planes of image * @param[in] mask_img pointer to dirt mask image * @param[in] dist_max threshold up to which dilation is done * @param[in] expand the dirt mask before replacing the pixels * @param[in] win_size size of adaptive mean filtering window * @param[in] smooth triangular filter whole image for grain removal * @param[in] inner find crop within or outside the image's edges * @param[out] crop array of 4 integers, if non-NULL, top, bottom, * left and right values for cropping are returned. * * @return * - SANE_STATUS_GOOD - success * - SANE_STATUS_NO_MEM - if out of memory * * The main purpose of this routine is to replace dirty pixels. * As spin-off it obtains half of what is needed for film grain * smoothening and most of how to crop positive film. * To speed things up these functions are also implemented. */ SANE_Status sanei_ir_dilate_mean (const SANE_Parameters * params, SANE_Uint **in_img, SANE_Uint *mask_img, int dist_max, int expand, int win_size, SANE_Bool smooth, int inner, int *crop); #endif /* not SANEI_IR_H */