Measure Utility

Image Comparison Utility

kalmus.utils.measure_utils.compare_needleman_wunsch(str_barcode_1, str_barcode_2, local_sequence_size=2000, match_score=2, mismatch_penal=- 1, gap_penal=- 0.5, extending_gap_penal=- 0.1, normalized=False)

Compare two input character arrays/strings (barcode)'s matching score using the Needleman Wunsch method. Needleman Wunsch: https://www.sciencedirect.com/science/article/abs/pii/0022283670900574?via%3Dihub

Parameters

• str_barcode_1 (str) -- The input string representation of barcode 1

• str_barcode_2 (str) -- The input string representation of barcode 2

• local_sequence_size (int) -- Divide the long barcode into several small barcode with local_sequence_size length

• match_score (int) -- The score (bonus) for correctly matching character

• mismatch_penal (int) -- The penalty for mismatch character

• gap_penal (int) -- The penalty for gaps within matched sequence

• extending_gap_penal (int) -- The penalty for extending gaps

• normalized (bool) -- If True normalize the final matching score into range [0, 1]. If False, return the raw score

Returns

The match score/normalized match score

Return type

float

kalmus.utils.measure_utils.compare_smith_waterman(str_barcode_1, str_barcode_2, local_sequence_size=2000, match_score=2, mismatch_penal=- 1, gap_penal=- 0.5, extending_gap_penal=- 0.1, normalized=False)

Compare two input character arrays/strings (barcode)'s matching score using the Smith Waterman method. Smith Waterman: https://www.sciencedirect.com/science/article/abs/pii/0022283681900875?via%3Dihub

Parameters

• str_barcode_1 (str) -- The input string representation of barcode 1

• str_barcode_2 (str) -- The input string representation of barcode 2

• local_sequence_size (int) -- Divide the long barcode into several small barcode with local_sequence_size length

• match_score (int) -- The score (bonus) for correctly matching character

• mismatch_penal (int) -- The penalty for mismatch character

• gap_penal (int) -- The penalty for gaps within matched sequence

• extending_gap_penal (int) -- The penalty for extending gaps

• normalized (bool) -- If True, normalize the final matching score into range [0, 1]. If False, return the raw score.

Returns

The match score/normalized match score

Return type

float

kalmus.utils.measure_utils.cross_correlation(signal_template, signal_source)

Signal matching. Cross correlation of two input signals. Signals need to be in the same shape

Parameters

• signal_template (numpy.ndarray) -- The template signal

• signal_source (numpy.ndarray) -- The source signal

Returns

The cross correlation between two input signals. High cross correlation means high similarity between two input signals. range in [-1, 1]

Return type

float

kalmus.utils.measure_utils.generate_brightness_string_from_brightness_barcode(brightness_barcode, num_interval=15)

Helper function Generate the string where each character represents the brightness interval of the brightness in the input brightness barcode.

Parameters

• brightness_barcode (numpy.ndarray) -- Input 1 dimensional brightness barcode with 1 channel. kalmus.barcodes.Barcode.BrightnessBarcode.brightness shape == [number of brightness, 1]

• num_interval (int) -- The number of intervals that will be divided in the brightness range [0, 255]

Returns

The string where each character represents the brightness interval of the brightness in the input

Return type

str

kalmus.utils.measure_utils.generate_hue_strings_from_color_barcode(color_barcode, num_interval=12)

Helper function Generate the characters strings that represent the hue values of the input RGB color barcode (3 channel in range [0, 255]).

Parameters

• color_barcode (numpy.ndarray) -- Input color barcode, the input barcode must be a 1 dimensional color barcode with kalmus.barcodes.ColorBarcode.colors three channels (R, G, B). shape == [number of colors, 3]

• num_interval (int) -- The number of intervals that will be divided in the Hue ring (0 to 360 degree)

Returns

The string where each character represent the hue interval of the colors in the input RGB barcode

Return type

str

kalmus.utils.measure_utils.local_cross_correlation(signal_template, signal_source, horizontal_interval=40, vertical_interval=40)

Local cross correlation between two input signals. The input signals need to be 2 dimensional for local windowing

Parameters

• signal_template (numpy.ndarray) -- The template signal

• signal_source (numpy.ndarray) -- The source signal

• horizontal_interval (int) -- Number of horizontal intervals (window width == signal width // horizontal intervals)

• vertical_interval (int) -- Number of vertical intervals (window height == signal height // vertical intervals)

Returns

The local cross correlation between two signals. Higher local cross correlation means higher similarity between two signals. range in [-1, 1]

Return type

float

kalmus.utils.measure_utils.nrmse_similarity(image_1, image_2, norm_mode='Min max')

Normalized root mean squared error (NRMSE).

Parameters

• image_1 (numpy.ndarray) -- The image 1 for comparison

• image_2 (numpy.ndarray) -- The image 2 for comparison

• norm_mode (str) -- The mode for the normalization, average mode use the max (||image_1||, ||image_2||) Min max use the max(image_1 value range, image_2 value range)

Returns

The score that measure the similarity between two images in range [0,1] using NRMSE 0 is the least similar, 1 is the most similar (same)

Return type

float

kalmus.utils.measure_utils.ssim_similarity(image_1, image_2, window_size=None)

Structural similarity index measure (ssim)

Parameters

• image_1 (numpy.ndarray) -- The image 1 for comparison

• image_2 (numpy.ndarray) -- The image 2 for comparison

• window_size (int) -- The size of the local window, integer

Returns

The Structural similarity index score in range [0,1] 0 is the least similar, 1 is the most similar (same)

Return type

float