Multi-DS Strategy for Source Camera Identification in Few-Shot Sample Data Sets

Algorithm 1

Selecting pseudo-label samples based on multiple distances.

	Symbols:
	: Set of unlabeled samples
	: Set of labels
	: Set of samples with label
	: Collection of distance measures
	: The distance threshold
	: The statistics threshold
(1)	Extract multiple feature vectors from all training samples;
(2)	Calculate multiple distance parameters as follows:
	for and do
	Form distance vector
	Sort in ascend order and take the first entries to form a new vector
	end for
(3)	Selecting pseudo-label samples:
	fordo
	fordo
	Count for in
	if Total count of is greater than a selected threshold then
	Label with pseudo-label (Note: one unlabeled may be marked with multiple pseudo-labels and appended to corresponding labeled sets)
	end if
	end for
	end for