Detecting significant events in consumer image collections

1. A method for determining significant events in a digital image collection, comprising, using a processor for:

a) generating image counts time-series from the image collection;

b) computing a model of the image counts time-series, wherein the model is an ARIMA(p,d,q) model; and

c) using the image counts time-series and the model to determine significant events.

2. The method of claim 1 , wherein the model is selected from a plurality of models using a goodness-of-fit score.

3. The method of claim 1 , wherein step c) includes

i) computing a set of residuals, wherein a residual is the difference between the image counts time-series and the image count predicted by the model at each time unit;

ii) identifying time steps of interest which have residuals above a computed threshold; and

iii) merging contiguous time steps of interest, and applying an image counts threshold to the merged contiguous time steps of interest to determine significant events.

4. The method of claim 1 , wherein the image counts time-series in step a) are generated at different time granularities.

5. A method for determining significant events in a digital image collection, comprising, using a processor for:

a) generating image counts time-series from the image collection at different time granularities;

b) computing a model of the image counts time-series at each time granularity, wherein the model is an ARIMA(p,d,q) model;

c) using the image counts time-series and the model to determine significant events at each time granularity; and

d) determining the significant events based on user actions or system requirements.

6. The method of claim 5 , wherein the model is selected from a plurality of models using a goodness-of-fit score.

7. The method of claim 5 , wherein step c) includes

i) computing a set of residuals, which is the difference between the image counts time-series and the image count predicted by the model at each time unit;

ii) identifying time steps of interest which have residuals above a computed threshold; and

iii) merging contiguous time steps of interest, and applying an image counts threshold to the merged contiguous time steps of interest to determine significant events.