Retroactive search of objects using

1. A method comprising, by a computing device:

at time t 2 , determining a delta set of first objects representing a difference between a first set of first objects at time t 1 and a second set of first objects at time t 2 , wherein each first object is a first image, wherein the first set of first objects is represented as a first k-dimensional tree, and wherein the second set of first objects is represented as a second k-dimensional tree;

for each first object, generating n elements, x 1 . . . x n , representing the content of the first object;

comparing the delta set of first objects with a set of second objects represented as a balanced k-dimensional tree, wherein each second object is a second image;

for each second object, generating n elements, y 1 . . . y n , representing the content of the second object, wherein when comparing the content of a second object with the content of a first object, the content of the second object substantially matches the content of the first object if

∑

i

=

1

n

⁢

(

x

i

-

y

i

)

2

is less than a pre-defined threshold; and

identifying each second object whose content substantially matches content of at least one first object from the delta set of first objects.

2. The method of claim 1 , further comprising

representing the set of second objects as the balanced k-dimensional tree, comprising:

for each second object, generating k elements representing the content of the second object;

partitioning the set of second objects into the balanced k-dimensional tree based on the k elements of each second object, comprising:

for each node at each level i of the k-dimensional tree,

if a sub-set of second objects belonging to a sub-tree associated with the node has one second object, then store the one second object in the node; and

if the sub-set of second objects belonging to the sub-tree associated with the node has multiple second objects, then

sort the sub-set of second objects according to their respective (i mod k)th one of the k elements;

storing a median second object in the node;

storing any second objects to the left of the median second object in a left sub-tree of the node; and

storing any second objects to the right of the median second object in a right sub-tree of the node.

3. The method of claim 1 , further comprising:

between time t 1 and time t 2 , receiving one or more new first objects; and

adding the new first objects to the first set of first objects to obtain the second set of first objects.

4. One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

at time t 2 , determine a delta set of first objects representing a difference between a first set of first objects at time t 1 and a second set of first objects at time t 2 , wherein each first object is a first image, wherein the first set of first objects is represented as a first k-dimensional tree, and wherein the second set of first objects is represented as a second k-dimensional tree;

for each first object, generate n elements, x 1 . . . x n , representing the content of the first object;

compare the delta set of first objects with a set of second objects represented as a balanced k-dimensional tree, wherein each second object is a second image;

for each second object, generate n elements, y 1 . . . y n , representing the content of the second object, wherein when comparing the content of a second object with the content of a first object, the content of the second object substantially matches the content of the first object if

∑

i

=

1

n

⁢

(

x

i

-

y

i

)

2

is less than a pre-defined threshold; and

identify each second object whose content substantially matches content of at least one first object from the delta set of first objects.

5. The media of claim 4 , wherein the software is further operable when executed to

represent the set of second objects as the balanced k-dimensional tree, comprising:

for each second object, generate k elements representing the content of the second object;

partition the set of second objects into the balanced k-dimensional tree based on the k elements of each second object, comprising:

for each node at each level i of the k-dimensional tree,

if a sub-set of second objects belonging to a sub-tree associated with the node has one second object, then store the one second object in the node; and

if the sub-set of second objects belonging to the sub-tree associated with the node has multiple second objects, then

sort the sub-set of second objects according to their respective (i mod k)th one of the k elements;

store a median second object in the node;

store any second objects to the left of the median second object in a left sub-tree of the node; and

store any second objects to the right of the median second object in a right sub-tree of the node.

6. The media of claim 4 , wherein the software is further operable when executed to:

between time t 1 and time t 2 , receive one or more new first objects; and

add the new first objects to the first set of first objects to obtain the second set of first objects.

7. A system comprising:

one or more processors; and

a memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to:

at time t 2 , determine a delta set of first objects representing a difference between a first set of first objects at time t 1 and a second set of first objects at time t 2 , wherein each first object is a first image, wherein the first set of first objects is represented as a first k-dimensional tree, and wherein the second set of first objects is represented as a second k-dimensional tree;

for each first object, generate n elements, x 1 . . . x n , representing the content of the first object;

compare the delta set of first objects with a set of second objects represented as a balanced k-dimensional tree, wherein each second object is a second image;

for each second object, generate n elements, y 1 . . . y n , representing the content of the second object, wherein when comparing the content of a second object with the content of a first object, the content of the second object substantially matches the content of the first object if

∑

i

=

1

n

⁢

(

x

i

-

y

i

)

2

is less than a pre-defined threshold; and

identify each second object whose content substantially matches content of at least one first object from the delta set of first objects.

8. The system of claim 7 , wherein the processors are further operable when executing the instructions to

represent the set of second objects as the balanced k-dimensional tree, comprising:

for each second object, generate k elements representing the content of the second object;

partition the set of second objects into the balanced k-dimensional tree based on the k elements of each second object, comprising:

for each node at each level i of the k-dimensional tree,

if a sub-set of second objects belonging to a sub-tree associated with the node has one second object, then store the one second object in the node; and

if the sub-set of second objects belonging to the sub-tree associated with the node has multiple second objects, then

sort the sub-set of second objects according to their respective (i mod k)th one of the k elements;

store a median second object in the node;

store any second objects to the left of the median second object in a left sub-tree of the node; and

store any second objects to the right of the median second object in a right sub-tree of the node.

9. The system of claim 7 , wherein the processors are further operable when executing the instructions to:

between time t 1 and time t 2 , receive one or more new first objects; and

add the new first objects to the first set of first objects to obtain the second set of first objects.