Workshop o mikrokontrolérech na SKSP 2024.
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

92 lines
3.3 KiB

4 months ago
# Configuration of the image library
######## General parameters #####################################################
ImageLib {
# Default tracing level (0 to disable)
Trace 0
# Limits for image allocation
ImageMaxDim 0xffff # Maximum width/height (at most 64k-1)
ImageMaxBytes 256M # Maximum size in bytes
}
#if CONFIG_IMAGES_DUP || CONFIG_IMAGES_SIM
######## Image signatures #######################################################
ImageSig {
# To find similar images, Sherlock uses comparison based on regions.
# First of all, the imagesim analyser extracts various region features.
# Sets of these features are called "image signatures" and they are stored
# in the `H' attribute of image objects.
# Signatures are later processed by the indexer to build an effective
# search structure finally used by the search server. See Indexer and Search
# sections for more options.
# Minimum image size to apply segmentation. Smaller images are always
# compared by the simple "average" method (see ImageSig.CompareMethod).
MinWidth 16
MinHeight 16
# List of subdivision thresholds in the first phase of segmentation.
# Lower the values to increase the average number of regions and vice versa.
PreQuantThresholds 6 12 15 20 25 25 30 30 40 40 50 50 60 60 60
# Settings for the second phase of segmentation -- usually not so important.
# We use an iterative algorithm to improve the average error from the first phase.
# We stop the process after PostQuantMaxSteps or if we get only PostQuantThreshold
# percentual improvement over the previous step.
PostQuantMinSteps 2
PostQuantMaxSteps 10
PostQuantThreshold 1
# BorderBonus and BorderSize parameters can increase or decrease the weight
# of image pixels near the borders. Weight of all pixels more than BorderSize * MIN(cols, rows)
# pixels far from the edges is 128. Then this value continuously decreses/increses up to 128+BorderSize.
BorderSize 0.4
BorderBonus -50
# Scaling constants for computation of normalized i-th order inertia features (I1, I2, I3).
InertiaScale 2 0.5 0.05
# Threshold for detecting textured images (see images/sig-txt.c for details).
# Decrease the threshold if you want less detected textures,
# set it to zero to disable the algorithm completely.
TexturedThreshold 0.32
# Signature comparison method:
#
# integrated
# based on: James Z. Wang, Jia Li and Gio Wiederhold,
# "SIMPLIcity: Semantics-Sensitive Integrated Matching for Picture Libraries",
# IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 23, no. 9, pp. 947-963, 2001.
#
# fuzzy (unstable and unbalanced parameters)
# based on: Yixin Chen and James Z. Wang,
# "A Region-Based Fuzzy Feature Matching Approach to Content-Based Image Retrieval",
# IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 9, pp. 1252-1267, 2002.
#
# average
# Simple distance of image features averages (ignores segmentation).
#
CompareMethod integrated
# Array of multiplicative constants in feature vector distance computation
# (L, u, v, LH, HL, HH, I1, I2, I3, X, Y). Each one must be an integer in range 0..15, default is 4.
CompareFeaturesWeights 4 6 6 4 4 4 4 4 4 4 4
}
#endif
#ifdef CONFIG_IMAGES_DUP
######## Duplicate finder #######################################################
ImageDup {
# Detection of image duplicates does not work yet.
}
#endif